WO2016200485A1 - Outage prevention in an electric power distribution grid using smart meter messaging - Google Patents

Outage prevention in an electric power distribution grid using smart meter messaging Download PDF

Info

Publication number
WO2016200485A1
WO2016200485A1 PCT/US2016/027754 US2016027754W WO2016200485A1 WO 2016200485 A1 WO2016200485 A1 WO 2016200485A1 US 2016027754 W US2016027754 W US 2016027754W WO 2016200485 A1 WO2016200485 A1 WO 2016200485A1
Authority
WO
WIPO (PCT)
Prior art keywords
message
messages
premises
multiplicity
outage
Prior art date
Application number
PCT/US2016/027754
Other languages
French (fr)
Inventor
Yinuo DU
Andrew Wright KIRBY
Adam David MERANDA
Original Assignee
Florida Power & Light Company
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Florida Power & Light Company filed Critical Florida Power & Light Company
Publication of WO2016200485A1 publication Critical patent/WO2016200485A1/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00Machine learning
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00Computing arrangements using knowledge-based models
    • G06N5/01Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06Electricity, gas or water supply
    • YGENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y04INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
    • Y04SSYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
    • Y04S10/00Systems supporting electrical power generation, transmission or distribution
    • Y04S10/50Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications

Definitions

  • the present disclosure relates to a system and method for utilizing smart meter messages to predict deficient elements of an electric power distribution grid prior to a customer reporting of an outage.
  • Electrical power distribution grids can be implemented as radial, loop or network type systems.
  • the distribution grids are arranged and interconnected to a substation in different ways depending on the type of system configuration. However for each type of distribution system configuration, the distribution circuits (commonly referred to as feeders and lateral feeders) distribute power delivered from the substation to loads at premises coupled to the grid through smart meters.
  • Various types of faults can occur in an electrical distribution system, some of which result in power outages, i.e., the loss of electric power service to customers.
  • a short circuit fault causes the protective element upstream of the fault to open isolating the short circuit fault from the grid.
  • a short circuit may be caused by a tree branch contacting power lines during a storm for example.
  • Customers downstream of the opened protective element become de- energized resulting in an outage.
  • Another type of fault is an open conductor element fault that similarly causes the downstream customers to experience a power outage.
  • An open conductor element may be caused by a power line snapping during a storm, or a coupling joining two power lines becoming deficient and then failing thereby resulting in the open conductor.
  • Power outage analysis conventionally relies on customer phone calls made to the utility company in the event of a power outage as the main information source for such analysis. This process can be quite slow because many customers may not call to report an outage, and those who do report an outage may wait a relatively long period of time to report the outage, often assuming a neighboring customer will call in their stead.
  • AMR automated meter reading
  • AMI advanced metering infrastructure
  • timelier message information is available from smart meters.
  • smart meters may not be designed to specifically detect deficient grid element and generate a corresponding message.
  • smart meters generate a multiplicity of message types, each of which has been shown to be an unreliable indicator of a service interruption that would qualify as an outage that requires a repair to an element of the power distribution grid.
  • the system and method of the present disclosure generally relate to the use of AMI messaging types and data mining decision trees to determine if local equipment failure is present.
  • the system and method may be used to predict impending failure and to create proactive investigation tickets.
  • the system and method can be applied to detect failures of higher level device equipment and may be incorporated into customer service processes (troubleshooting during first customer contact). Further, the system and method may also be used to determine customer owned equipment failures for referral to electricians (inside the premises trouble).
  • FIG. 1 illustrates one implementation of a method to create a Single premises No Current (SNC) ticket based on a predictive algorithm.
  • SNC Single premises No Current
  • FIG. 2 illustrates one implementation of a method to investigate single no current prediction tickets created through the process illustrated in FIG. 1.
  • FIG. 3 illustrates the monitoring of AMI smart meter messages to ascertain the likelihood of an SNC outage in accordance with one implementation.
  • FIG. 4 illustrates a graphical user interface for the premises troubleshooting tool in accordance with one implementation.
  • FIG. 5 illustrates a chart showing last gasp messages observed in order to ascertain whether a true sustained power outage has taken place in accordance with one implementation.
  • FIG. 6 illustrates a chart that reflects intermittent power conditions in accordance with one implementation.
  • FIG. 7 illustrates at a high level the flow of information from a smart meter in accordance with one implementation.
  • FIG. 8 illustrates a methodology for selecting power failure prediction models in accordance with one implementation.
  • FIG. 9 illustrates a user interface that may be used to display notifications alerting a utility that a power failure is likely to occur in accordance with one implementation.
  • FIG. 10 illustrates an exemplary set of event messages retrieved from an AMI database in accordance with one implementation.
  • FIG. 11 illustrates a combination of ticket and interruption types that may be analyzed in one implementation.
  • FIG. 12 illustrates an exemplary histogram or time series plot of AMI event messages in accordance with one implementation.
  • FIG. 13 illustrates a sample of the data prepared for input into a data mining engine in accordance with one implementation.
  • FIG. 14 illustrates a sample screenshot of a data extraction process in accordance with one implementation.
  • FIGs. 15A-C illustrate a decision tree that may be used to predict power failures in accordance with one implementation.
  • FIG. 16 is a simplified block diagram illustrating components of a metered electric utility distribution system.
  • the systems and methods described herein may be implemented as a method, data processing system, or computer program product. Accordingly, the description may take the form of an entirely hardware embodiment, an entirely software embodiment, an embodiment combining software and hardware aspects, a computer program product on a computer-usable storage medium having computer readable program code on the medium, a non-transitory computer readable storage medium, or combinations thereof. Any suitable computer readable medium may be utilized including, but not limited to, static and dynamic storage devices, hard disks, optical storage devices, and magnetic storage devices.
  • the disclosed system and method may be implemented on one or more computing devices, including one or more servers, one or more client terminals, including computer terminals, a combination thereof, or on any of the myriad of computing devices currently known in the art, including without limitation, personal computers, laptops, notebooks, tablet computers, touch pads (such as the Apple iPad, SmartPad Android tablet, etc.), multi-touch devices, smart phones, personal digital assistants, other multi-function devices, stand-alone kiosks, etc.
  • computing devices including one or more servers, one or more client terminals, including computer terminals, a combination thereof, or on any of the myriad of computing devices currently known in the art, including without limitation, personal computers, laptops, notebooks, tablet computers, touch pads (such as the Apple iPad, SmartPad Android tablet, etc.), multi-touch devices, smart phones, personal digital assistants, other multi-function devices, stand-alone kiosks, etc.
  • FIG 16 illustrates a simplified block diagram of elements of an electric power distribution system.
  • the distribution system 100 includes a power generation source 110 that is operative to generate electric power in accordance with one or more technologies and to distribute the electric power via transmission lines 120.
  • source 110 may employ or included a hydro-electric plant, a fossil fuel or coal-fired plant, a nuclear plant, a wind farm or a solar array or collector and attendant structures, a geothermal plant, or a combination of these and other facilities that are known or developed in accordance with known principles and technologies to generate electric power.
  • transmission lines 120 may carry high- voltage Alternating Current (or "AC") electric power from source 110 to a desired destination as is generally known, though other technologies for distributing electric power across distances, as may be appropriate for a particular application and based upon technological readiness and suitability, are contemplated.
  • AC Alternating Current
  • the present disclosure is not intended to be limited by the structural or operational characteristics of source 110 or transmission lines 120, or by the nature of the electric power and the manner in which it is generated or transmitted.
  • electric power generated at source 110 may be transmitted via transmission lines 120 to a power transfer substation 130 which, in turn, distributes electric power to transformers (Txi, Tx 2 , . . . Tx n ) 140-141 for subsequent transmission to a multiplicity of premises or a group of premises 144-145 having a corresponding group of smart meters that are generally represented as meters (Mi, M 2 , . . . M n ) 150-151.
  • electric power distribution system 100 may generally include more power generation sources (such as source 110) and more substations (such as substation 130) than depicted in FIG. 16, and that each such substation 130 may serve many more transformers (such as transformers 140) than illustrated.
  • each such transformer 140 such as specific transformer 141, may provide electric power to more premises and smart meters (such as meters 150-151) than are depicted in FIG. 16.
  • the present disclosure is not intended to be limited by the simplified topography represented in the drawing figures.
  • substation 130 may modify or condition the electricity received from source 110 such that it may then be transmitted to transformers 140. For example, it may be desirable in some circumstance to step down (or to step up) voltage via one or more substation transformers 131, or to phase-shift or otherwise to adjust current phase or amplitude, for instance, to achieve a desired power function as specified by the kind of load and/or to minimize energy lost in the distribution system.
  • steps down (or to step up) voltage via one or more substation transformers 131, or to phase-shift or otherwise to adjust current phase or amplitude, for instance, to achieve a desired power function as specified by the kind of load and/or to minimize energy lost in the distribution system.
  • phase-shift or otherwise to adjust current phase or amplitude, for instance, to achieve a desired power function as specified by the kind of load and/or to minimize energy lost in the distribution system.
  • Various techniques are known or may be developed to condition electric power at substation 130, and the present disclosure is not intended to be limited by the operation of substation 130 or by any technical
  • transformers 140-141 may be configured and operative further to condition the electric power received from substation 130 such that it is suitable for delivery to customers or loads associated with smart meters 150-151.
  • Voltage manipulation, current manipulation, or a both may be employed in various situations; the nature and extent of such conditioning may be dependent upon the specifications and operational characteristics of the smart meters 150-151 to which electric power is delivered, for example, or upon governmental regulations, technological or infrastructure capabilities, or a combination of these and other factors.
  • the present disclosure is not intended to be limited by the technologies employed at transformers 140-141.
  • Smart meters 150-151 may be configured to measure electric power consumed at each corresponding premises and generate a multiplicity of message types related to meter electric power usage at a particular premises such as a residence, building, business location, or some other specific site. In some circumstances (where available from and supported by a particular utility service provider, for instance, or where required by a state or local utility oversight commission or committee), meters 150-151 may be operated in accordance with Advanced Metering Infrastructure (or "AMI") specifications or protocols.
  • AMI Advanced Metering Infrastructure
  • meters 150-151 may be configured as or include what are generally referred to as “smart meters” or “AMI meters.”
  • meters 150-151 may monitor, sense, record, or otherwise track electrical consumption (or “usage") digitally and at predetermined or dynamically adjustable intervals; subsequently or effectively contemporaneously, meters 150-151 may transmit usage data (e.g., using a secure radio frequency band or other telecommunications methodology) to other devices installed on or otherwise communicably coupled to the electrical grid or to a communications network (such as AMI network 160) with which meters 150-151 may exchange data.
  • usage electrical consumption
  • meters 150-151 may transmit usage data (e.g., using a secure radio frequency band or other telecommunications methodology) to other devices installed on or otherwise communicably coupled to the electrical grid or to a communications network (such as AMI network 160) with which meters 150-151 may exchange data.
  • AMI network 160 such as AMI network 160
  • usage data may refer to raw data (i.e., unmodified or unprocessed data as they are collected and that represent the amount of electrical energy consumed at the location metered by meters 150-151) or to data and other information relating to or derived from such raw data (e.g., readings or data points that may be time-stamped or otherwise processed to provide more information than simply an aggregate or a measure of cumulative consumption).
  • an aggregate usage value may be transmitted from such a smart meter, while in other instances, detailed, time-dependent usage rates may be transmitted; the specific type and amount of data collected and processed at, and transmitted from, meters 150-151 may be application-specific and may vary in accordance with processing or computational capabilities of hardware components deployed in, as well as software functionalities implemented at or in cooperation with, meters 150-151. These factors may be affected by technological or economic considerations, for example, or may be dictated or influenced by applicable governing bodies or governmental regulations.
  • smart meters such as meters 150-151 may be communicably coupled to AMI network 160 such that messages including usage data collected (and possibly processed) by meters 150-151 may be transmitted to AMI network 160.
  • AMI network 160 may be communicably coupled to a utility provider's facilities and computer systems (hereinafter generally referred to as "systems" 180).
  • the AMI network may include the Internet, for instance, or any other packet-switched data network, virtual private network, proprietary or public wide area network (“WAN”), or any other communications network capable of bi-directional data communication between AMI network 160 and systems 180.
  • WAN wide area network
  • systems 180 generally include processing and billing systems, various monitoring, customer service, troubleshooting, maintenance, load balancing, accounting, and other types of activities that may be used to operate a utility service. In some instances where such activities are computationally expensive or require a great deal of processor power or communication bandwidth, some implementations of processing may be distributed across many processing and memory storage resources, and even distributed across buildings as is generally known in the art. Accordingly, though systems 180 are depicted in FIG. 16 as a single, unified entity, it will be appreciated that block 180 may represent any number of physical buildings, computer server facilities, and other physical and logical resources owned, operated, or otherwise under the control of the relevant utility service provider.
  • a customer such as a customer located at premises 1, 144 may experience an interruption in service, or an outage.
  • An outage may result from an electric power distribution grid having any of a multiplicity of elements becoming deficient of failing.
  • the elements include any component of the grid coupling power generation source 110 to premises 144, including power generation source 110. If an element between or including power generation source 110 and transformer 140 becomes deficient and fails, then a multiple premises outage may result.
  • a single premises outage may occur if an element between transformer 151 and premises 144 becomes deficient or fails.
  • a single premises outage may occur for any of a multitude of reasons.
  • transformer 142 has a power coupling 250 for providing power to premises 1 through n, 144-145. If a portion of power coupling element 250 becomes deficient, such connection corrosion or insulation deterioration, then an outage at premises 144 may occur. If power distribution line 252 becomes deficient, either by corrosion or high impedance leakage or a fault in an underground conduit, then an outage at premises 144 may occur.
  • PSTN public switched telephone network
  • AMI network 160 used for communicating smart meter messages from smart meter 150 at premises 144 to utility 180, even though various components may be shared between networks, such as internet or cellular networks.
  • the PSTN network 190 is used for customer communications with the utility and the AMI network 160 is used for smart meter message communication with the utility.
  • the PSTN and AMI networks operate independently and facilitate different types of communication.
  • alternate communication devices other than phone 258, may be utilized to communicate an outage from a customer to the utility while remaining within the scope of this description. Other devices may include smart phones, personal computers or other devices for customers to communicate with the utility.
  • outage processor 260 In response to an outage report, outage processor 260 process the outage report and determines an appropriate response. Ticket generator 270 then generates a ticket via a ticket output device in the form of a humanly observable repair ticket (such as a printed page or display on a computer screen) indicative of the prediction, thereby enabling a repair crew 275 to repair the deficient grid element associated with the outage at the premises. The findings of the repair crew may then add to the outage report.
  • the outage processor may use a number of approaches known to those familiar in the art for determining an appropriate response. Such approaches may include analyzing a multitude of outage reports from a multitude of premises. In order to keep customer satisfaction high and to reduce maintenance costs, it is desirable to reduce the amount of tickets and the amount of repair work associated with the grid.
  • Ticket type represents what the customer reported via the phone system and before performing a field investigation include:
  • SNC(U) single no current (overhead or underground), used to report PNC conditions, and
  • NLS no loss of service, used to report power quality or other field conditions which cannot be deferred to regular work (reachable low wire, for example).
  • Interruption type represents the category of the outage after the field investigation and repairs are complete and include:
  • OTH-SEC multiple customer service failure (including failures of handholes or other junctions downstream of a distribution tx),
  • FDRs which may be defined as main power lines traveling out of a substation
  • LATs laterals
  • SUBs substations
  • TXs transformers
  • single premises events may be responsible for less than 3% of the customer outages while generate over 74% of the repair ticket volume, thereby making up the majority of restoration work of an electric power distribution grid utility's maintenance fleet.
  • Single premises events include single no current (SNC) tickets, which may relate to a single customer or meter reporting a loss of power, no loss of service (NLS) tickets, which may relate to an outage ticket type generated when a customer is reporting that there is no sustained outage or interruption of electric service at the time, and location (LOC) tickets, which may relate to an outage ticket type where the suspected device cannot be identified.
  • SNC single no current
  • NLS no loss of service
  • LOC location
  • single premises events result in high number of tickets. Reducing single premises events may significantly improve the quality of service provided to customers of a utility providing electric power to customers. Consequently, it may be advantageous to predict the occurrence of single premises events before an outage is reported.
  • a "bad truck roll” may be defined as a situation when either no utility problem exists or an investigator is unable to determine the problem and yet the investigator makes a trip to a premises or location to conduct an investigation based upon a ticket.
  • One potential advantage of the present disclosure is to reduce investigator volume.
  • a single premises event or outage may typically result from a failure related to grid elements 250-256.
  • Smart meter 150 is able to generate a message as a result of an outage, such as a last-gasp message.
  • reception of a single last gasp message (or multiple last gasp messages generated by a single smart meter) by AMI processing function 189 is typically not reliable enough to substitute for an outage report generated in response to a customer call.
  • elements of the power grid such as elements 250-256 become deficient slowly before a failure results in an outage. Power couplings may slowly corrode before failure, insulators may slowly become contaminated before failure, underground cable may slowly degrade before failure, vermin may slowly destroy elements before failure.
  • smart meters While smart meters generate a multiplicity of messages, the messages do not include information specifying a deficient grid element. For example, smart meter messages do not specify deficient power couples 250 or 254, nor do they specify deficient lines 252 or 256.
  • the inventors have discovered how to determine a smart meter message behavior that is indicative of deficient grid elements and predictive of future outages, and then apply the smart meter message behavior to messages received by smart meters in the grid to identify a suspect premises and generate a ticket to enable a repair crew to begin investigation and/or repair a deficient grid element prior to an outage being reported by a customer.
  • dispatching a repair crew to a premises prior to receiving an outage call is also an improvement in customer service, even though the customer had to call the utility, the utility can respond that a deficient element was detected and a repair crew has already been dispatched prior to the customer placing the call.
  • FIG. 16 shows An electric power distribution system 100 comprising an electric power distribution grid having a multiplicity of elements 250-256 capable of becoming deficient, a multiplicity of smart meters 150-151 coupling the electric power distribution grid and a corresponding multiplicity of premises 144-145, each smart meter measuring electric power consumed at each corresponding premises and generating a multiplicity message types related thereto, the multiplicity of message types not including information specifying a deficient grid element.
  • FIG. 16 shows An electric power distribution system 100 comprising an electric power distribution grid having a multiplicity of elements 250-256 capable of becoming deficient, a multiplicity of smart meters 150-151 coupling the electric power distribution grid and a corresponding multiplicity of premises 144-145, each smart meter measuring electric power consumed at each corresponding premises and generating a multiplicity message types related thereto, the multiplicity of message types not including information specifying a deficient grid element.
  • FIG. 16 also shows an automated meter infrastructure network 160 coupled to the smart meters for communicating the messages from the smart meters and a message database within AMI processor 189 coupled to the automated meter infrastructure network for receiving messages from the multiplicity of smart meters, and an outage processor 260 for receiving outage reports through a communication network 190 different from the automated meter infrastructure network 160, each outage report associated with a failed deficient grid element.
  • FIG. 16 also shows an outage predictor 265 for processing the additional messages with the plurality of models to predict a likelihood of an outage at a suspect premises based upon messages from smart meters, and a ticket generator 270 for generating a ticket enabling a repair crew to repair a deficient grid element associated with the suspect premises.
  • a data mining process may be applied to messages having event data transmitted by smart meters.
  • a data mining process may be defined as a process of analyzing data from many different perspectives, which may allow, for example, viewing information from many different perspectives or "angles" at the same time. For example, instead of sampling data points, the entire population of available data may be used to find patterns predictive of outages.
  • the process data mines smart meter messages to look for patterns indicating whether a future single premises interruption may be likely to occur.
  • one of the methods of the present implementation may rely on event correlation.
  • Meters send structured messages, which include event information, over the AMI network.
  • a history of smart meter messages received over the AMI network and a corresponding history of outage reports received over the PSTN network may be analyzed.
  • a process receives these structured smart meter messages and filters out duplicate messages.
  • Duplicate messages may be received since the same message generated at a meter may be routed through different nodes in a mesh based AMI network.
  • the messages and outage report database may be organized into a data structure.
  • the data structure associated with the meter messages may relate to many attributes of the meter. For example, an AMI meter may generate over 280 types of event messages.
  • the data mining engine may be IBM's SPSS tool (http://www- 03.ibm.com/software/products/en/spss-decision-trees).
  • a data mining engine may be implemented as statistical analytical software used for data analysis, data mining and/or forecasting. Use of the data mining engine may enable users to import (from a variety of sources including databases, flat files, IBM Cognos, Excel, XML, etc.), manage, and analyze data in a graphical user interface. It may also allow the user to construct graphs and charts, tables, cross- tabulations, build forecasting models, decision-trees, and perform a number of advanced statistical functions and export them to the desired format.
  • the data mining engine may generate equations describing message behaviors based on any of a number or theories including for example, a natural theory of entropy. These equations or algorithms may then be used to predict outages.
  • the equations are represented as a visual and textual breakdown of a decision tree. The decision tree is selected based on a comparison of different correlations among meter events and selecting the decision tree which best predicts failures.
  • the decision tree analysis may be used to model a series of events and look at how they affect an outcome by calculating a set of conditional probabilities based on different scenarios.
  • a CHAID (CHi-squared Automatic Interaction Detection) decision tree may be selected for failure prediction.
  • a different process may be selected for splitting and pruning a decision tree, and noise handling may be selected, such as C5.1 or other decision tree process known to those skilled in the art.
  • algorithm logic may be defined and delivered. The algorithms are capable of generating prediction models for predicting single premises outages before they occur based upon messages from smart meters. The final algorithms look for groups of event messages in a hierarchical method.
  • application of the decision trees shows that premises with SNC (interruption type SEC, SVC) have smart meters that send certain types of messages or message behaviors at a rate or occurrence or in certain concentrations 48-24 hours before an outage occurs.
  • the process may generate a NLS ticket, then upgrade to a SNC upon receiving customer call reporting an outage.
  • the process can help reduce the number of "bad truck rolls.”
  • FIG. 1 and FIG. 2 describe a use of the message behaviors for implementing corrective actions.
  • FIG. 1 illustrates one implementation of a method to create a SNC ticket based on a predictive algorithm using smart meter message behavior.
  • the process opens a list of flagged premises.
  • the premises may be flagged by an outage predictor which flags premises based on at least an output smart meter messages processed by any of a plurality of decision tree based models.
  • the process filters for premises flagged in last 24 hours.
  • the process filters premises for current area of interest.
  • the process determines whether there are any new premises flagged, and if no new premises are flagged, then in step 109 the process closes a list of flagged premises.
  • the process may optionally go on to notify a team, such as an evaluation team, of the results.
  • step 111 determines whether a premises had a ticket in the last 48 hours. If the premises had a ticket in the last 48 hours, then in step 113 the process monitors the premises for another 48 hours, and then the process in step 115 determines whether the premises is continuing to be flagged, and if the premises does not continue to be flagged, the process ends.
  • step 117 the process opens a premises troubleshooting tool (see Fig. 4) after it has been determined that a premises did not have a ticket in an initial period of 48 hours (step 111) or after it has been determined that a premises continues to be flagged after being monitored for an additional 48 hour period following the initial 48 hour period (steps 113 and 115).
  • the premises troubleshooting tool allows selection of a date range and assignment of a premises number in step 119.
  • Step 121 records premises information;
  • step 123 records daily average kwh usage;
  • step 125 records meter ping result (e.g., ping for instantaneous voltage data);
  • step 127 records event summary details; and step 129 opens a ticket summary tab.
  • step 131 the process determines if a new NLS ticket is required, and if no new NLS ticket is required the process ends. If a new NLS ticket is required, then the process, in step 133, opens TCMS and creates a new trouble call management system ("TCMS") ticket including the NLS ticket. In step 135 the process adds inspection instructions and note "Technology" to ticket thereby indicating that the ticket is created by the utility predictive model and not a conventional customer outage report. In step 137 the process records details and comments on the flagged premises list. In optional step 139 the process notifies a team, such as the pilot team, of the results.
  • TCMS trouble call management system
  • FIG. 2 illustrates one implementation of a method to investigate single no current prediction tickets created through the process illustrated in FIG. 1.
  • the process receives an NLS prediction ticket generated in FIG. 1.
  • the process investigates the ticket as shown on the restoration process for restoring power (which is part of a standard repository of business processes: RST— 0.6).
  • the process generates a signal for notifying the customer of proactive equipment failure evaluation details.
  • the ticket is sent to a repair crew with instructions to follow for the repair.
  • the ticket includes extra instructions from the trouble call management system ("TCMS").
  • the extra instructions may be included in the customer remarks field.
  • the TCMS extra instructions may be used to enhance the management of outages or identification of potentially defective grid elements.
  • a TCMS ticket may be sent to a crew with specific steps that let crew know what to do in the event of intermittent power problems such as conduct additional due diligence, follow through extra steps and document in the system.
  • the process determines whether there is a deficient grid element resulting in equipment failure present based upon the finding or investigation of the crew. If there is no equipment failure the process determines whether the ticket needs to be referred for equipment repair (step 215). If the ticket does not need to be referred, the ticket is closed (step 217).
  • step 209 If there is equipment failure (step 209), then as part of the process of the investigation of the crew, a photograph may be taken (step 211) of the equipment failure, and is attached to the ticket, and in optional step 213 the customer of a proactive equipment failure investigation result is notified.
  • step 215 If a ticket indicating a deficient grid element needs to be referred for repair (step 215), then in step 219 the deficient grid element equipment is repaired.
  • the equipment may be repaired in accordance with a standard business process which may be described for example in restoration process RST-0.6.
  • step 221 the process determines whether the damaged equipment was removed during repair. If the damaged equipment was not removed, then the process in step 223 updates the ticket as shown on restoration process RST— 0.6 and the ticket is closed.
  • step 225 the process quarantines defective equipment.
  • the quarantined equipment may be used by an evaluation team to further enhance the failure prediction model. For example, damaged cables may be removed and held on to so that the utility can see the damage and add the results to the prediction model generator.
  • the process updates (step 227) the ticket as shown in restoration process RST 0.6, and the ticket is then closed (step 237).
  • FIG.3 illustrates a decision tree based model for the monitoring of AMI smart meter messages to ascertain the likelihood of an SNC outage in accordance with one implementation. It has been discovered that premises with SNC outages exhibit a message behavior that results in the smart meter sending certain types of messages in certain concentrations 24-48 hours before the SNC outage.
  • NIC Network Interface Card
  • Service Error Cleared which is a type of AMI event indicating that there was an issue detected by the meter but that it has cleared itself
  • FIG. 4 illustrates a graphical user interface ("GUI") produced by the premises troubleshooting tool in accordance with one implementation.
  • This tool may be used for investigating a flagged premises.
  • the GUI 413 prompts the user for premises information, which is then uploaded into the tool.
  • the GUI displays detailed premises information 401 for a premises that has been flagged by the process of the present disclosure, Event Summary information 403, detailed Event information 405, and ticket information 407.
  • the GUI also displays a graph plotting data corresponding to power usage in kWh and corresponding to measured voltage in volts root mean square (Vrms).
  • Vrms volts root mean square
  • a voltage and usage anomaly 409 caused the premises to be flagged.
  • the anomaly shows the voltage and power dropping to zero after September 2.
  • the GUI also displays summary voltage and usage information 411.
  • FIG. 5 illustrates a chart 600 showing last gasp messages which do not represent a true sustained outage.
  • a last gasp message may be defined as an event generated when an AMI meter detects a zero voltage event for a period of time (e.g., 45 seconds). The voltage and power at a premises are shown dropping to zero after June 11. After the drop numerous smart meter messages including last gasp messages are generated as represented by dots occurring after June 11. While focusing on the valuable messaging from the AMI network chatter may yield important information about outages, because the amount of last gasp messages illustrated in FIG. 5 does not represent a true sustained outage, the last gasp messages alone are not entirely accurate for generating reliable single premises tickets.
  • FIG. 6 illustrates a chart 700 that reflects intermittent power conditions. In one embodiment, the processes for predicting power failure exploit smart meter message patterns with intermittent or partial power problems.
  • FIG. 7 illustrates at a high level the flow of information (notifications, voltage reads, event messages, etc.) from a meter.
  • each smart meter generates a significant number of voltage read messages and significantly more event messages.
  • over one point two million event messages across over two hundred and eighty categories or message types may be generated in a twenty four hour period while only sixty single premises outages may occur within the period.
  • processing one point two million messages per day to predict sixty outages before a customer generate outage report occurs may be akin to finding a preverbal needle in a haystack.
  • a typical utility may have very many more smart meters in the power distribution system, over four million smart meters for example.
  • the resulting very large amount of data may be stored in a database.
  • the database may be a database adapted for storing big data files in a big database.
  • Such a database may be an extensible relational database platform using a massively parallel processing approach such as a Greenplum database, as is known to those familiar with the art.
  • Historical data may be housed in the database and then analyzed using data mining techniques in accordance with the present disclosure.
  • These notifications, reads, and event messages may include last gasp messages, power failure count messages, NIC power down messages, inversion count notifications, zero voltage messages, DC detection cleared messages, and other types of messages as would be recognized by those skilled in the art.
  • Numeral 801 represents the database that contains meter event data.
  • Numeral 805 represents a step of scrubbing meter event data 803 collected over a two day period. The data is scrubbed for duplicates, errors, etc.
  • the data is loaded to a data mining engine 807.
  • the data mining engine considers meter event message information as well as customer generated power outage information.
  • different models may be created 809 for the prediction of power failures.
  • two decision tree models, such as CHAID and C5.1 are selected such that event data can be fed into the algorithms and power outages can be predicted before being reported by a customer.
  • the Prediction Algorithm rectangle of FIG. 8 shows an example of an algorithm indicative of a message behavior based upon a branch of a decision tree model, e.g., ⁇ message 1+ (message 2 ⁇ X% concentration) OR (message 2 > X %+ message 3 ⁇ Y concentration)) ( message 4 ⁇ Z % concentration) ⁇ .
  • FIG. 9 illustrates a user interface that may be used to display notifications 901 alerting the utility of a prediction that a power failure is likely to occur at a suspect premises.
  • the user interface may display premises information 909, voltage readings 907, meter event summary 903, and a history of events 905.
  • the aforementioned decision trees are applied to the event data in order to ascertain whether to display an alert 901.
  • the information displayed in FIG. 9 including notification 901 may allow a ticket generator to generate a ticket enabling a repair crew to repair a deficient grid element associated with the suspect premises. Given the predictive nature of the notification the repair may occur prior to a failure of a defective grid element, or prior to a generation of an outage report.
  • FIG. 10 illustrates an exemplary set of event messages retrieved from an AMI database (e.g., a GreenPlum database).
  • AMI database e.g., a GreenPlum database
  • event messages from a time period may be selected for extraction.
  • more than 100 million event messages may be extracted and saved in a CSV file.
  • PostGreSQL is one of the possible tools known to those familiar with the art that may be used to access a GreenPlum database.
  • AMI meter event data or information may be extracted from GreenPlum:
  • a complex data cleanup may then be started.
  • the data cleanup may include defining and removing duplicate messages and messages from commercial industrial meters or other meters not associated with a single premises.
  • a duplicate message may be defined as any message which has been sent more than once from the same meter within a time period, such as one second.
  • the process may query all single premises trouble or outage tickets from an outage processor which may be included in a Power Delivery Distribution data warehouse. Premises which experienced an outage during the time frame of interest may be noted, and a column added from the outage processor to the corresponding data record to indicate whether an event message was sent by a meter before (which may be defined as any event which occurred more than 6 hours before the meter was interrupted), during (which may be defined as any event from 6 hours before the meter was interrupted until power to the meter was restored), or after (which may be defined as any event which occurred after power was restored to the meter) an outage event.
  • a meter before which may be defined as any event which occurred more than 6 hours before the meter was interrupted
  • during which may be defined as any event from 6 hours before the meter was interrupted until power to the meter was restored
  • after which may be defined as any event which occurred after power was restored to the meter
  • FIG. 11 illustrates a combination of ticket and interruption types that may be analyzed in one implementation.
  • FIG. 11 shows eight columns with corresponding ticket types and seven rows with corresponding interruption types as analyzed by a repair crew in response to a ticket.
  • tickets which started as a single premises interruption (single no current - SNC/single no current underground - SNCU) and ended as a secondary (SEC) or service (SV) interruption (e.g., service indication of intermittent problems) are considered as part of the algorithm development for predicting failures.
  • SEC secondary
  • SV service interruption
  • a time series analysis or study may be undertaken of the data set resulting from the selection of FIG. 11.
  • all event messages for meters which experienced an outage may be listed from oldest to newest in order to facilitate review of when the messages were sent.
  • the time series analysis reveals that most event messages occur in a time period preceding an outage.
  • FIG. 12 illustrates an exemplary corresponding histogram or time series plot of AMI event messages.
  • the histogram of FIG. 12 shows a frequency of twelve event or message types occurring between twenty five hours before and five hours after an outage. In this example, the frequency of events increases significantly before an outage.
  • messages occurring within a period ranging from 54 hours before to 6 hours before an outage may be evaluated. The difference between the raw count of events in that 48 hour period may be compared to a corresponding 30 day historical average.
  • all AMI events stored in the GreenPlum database and occurring outside of any chosen time frame may be considered as irrelevant for purposes of creating a model for failure prediction.
  • a routine extract may be set up to query the total number of events occurring in the last 24 to 48 hours for comparison against a long term historical average.
  • the query may be run under the assumption that it will be 50% successful, meaning that 50% of meters which sent an above average (>2x) rate of event messages are likely to experience an outage.
  • the data may be imported into the data mining engine (for example, IBM's SPSS modeler software). Once inputted into the data mining engine, the data may be transformed such that the event message concentration for each AMI smart meter is calculated. The calculation is shown in equation 1.
  • IBM's SPSS modeler software for example, IBM's SPSS modeler software.
  • Equation 1 Calculating the Concentration of Meter Messages
  • the message behavior includes a ratio of first messages (15 NIC Power Down messages) over second messages (100 Total Messages), the first messages including a first message type and the second messages including the first message type and at least a second message type, the ratio being 15% which is greater than zero and less than one. Also, it was determined that only 82 events were consistent with electrical messaging, and so the Greenplum AMI event data set was reduced to calculate the message concentrations for these event types.
  • FIG. 13 A sample of the data prepared for input into the data mining engine is illustrated in FIG. 13.
  • an auto-classifier function may be applied to the data, which can be used to determine which models perform the best prediction of a targeted condition.
  • column A shows the target condition to be an interruption or outage.
  • Column B shows the premises number (in scientific notation) and the remaining columns show twelve of the eighty two event types consistent with electrical messaging and relevant to the outage.
  • Each of the cells associated with the eighty two event types shows a percentage of the concentration of events (shown as a whole number).
  • prediction models may be selected.
  • a CHAID Cho Squared Automatic Interaction Detection
  • a C5.1 model may be selected for application to real-time data being received from the meters and predict power failures.
  • the models may be selected based on the analysis of 100 failures in the data set and more than 4.6 million meters without a failure condition.
  • 30 or more replicate models may be created and analyzed which may be generated from C5.1, Log Regression, CHAID or other modeling algorithms.
  • the output from the data mining engine may be a decision tree, so, in the example implementation described 30, decision trees from two separate modeling algorithms (C5.1 and CHAID) may be compared for consistency.
  • Equation 2 Predictive Algorithm Event Message Concentration ⁇ NIC Based Power Fail Detect Disabled + (History of Direct Current Detected ⁇ 0.1% concentration +) OR (History of Direct Current Detected >0.1% + NIC Power Down ⁇ 12.5% concentration)) ( Service Error Cleared ⁇ 1.5% concentration) ⁇ [0088]
  • a set of SQL queries may be applied so that an automatic data extract of premises meeting the algorithm conditions is executed at 15 minute intervals, in accordance with one embodiment.
  • a sample screenshot of the extraction process is shown in FIG. 14. The results of these queries may be stored in a SQL database where they can be retrieved on demand or according to a script.
  • FIG. 15A-FIG. 15C illustrates a decision tree that may be used to predict power failures.
  • the tree includes tree nodes (as opposed to network nodes) 1501 to 1551.
  • Each node may be associated with a type of event message that is considered to be relevant. The higher the messages appear in the tree structure, the more relevant the message will be in connection with predicting a failure.
  • the predicted outage or failure is associated with a failure that will result in the generation of a PNC ticket.
  • the following event messages are illustrated: NIC B.P.F., NIC P.D.C.U., Zero V.R., Last Gap, P.F.C., NIC P.D., and NIC F.T.S.
  • each node may include a node identification number (for example "Node 1"); a category (for example, "1” means a failure and "0" a no failure); a percentage of confidence level; and the number of premises "n” that had a particular event.
  • the p-value is the measurement of statistical significant level
  • the chi-square value represents calculated statistics based on chi-square equation
  • the "df ' is the degree of freedom - the number of values in the final calculation of a statistic that are free to vary.
  • An exemplary "split" is described with respect to nodes 1501 to 1505.
  • the number of premises or locations associated with node 1501 is 10,100 (9999 plus 101).
  • 47 are identified in node 1503 as having sent a NIC B.P.F. event message (indicated by the > 0.000 notation).
  • the confidence level is 100% - a customer call was generated in connection with all of the 47 premises to report an outage. Therefore, the system of the present disclosure can be used to predict failures before they occur by processing the additional smart meter messages the model of node 1503 to predict a likelihood of an outage at a suspect premise.
  • the confidence level threshold can be set to be high, for example, 75% or more, in order to achieve a desired predictability level.
  • nodes 1503, 1509, 1513, 1523, and 1535 include a confidence level of over 75% in connection with category "1" failure events. These nodes are selected from the multiplicity of end nodes of FIG. 15 (which produce a corresponding multiplicity of prediction models) thus produce selected message behaviors (from a plurality of prediction models 1503, 1509, 1513, 1523, and 1535 selected from the multiplicity of prediction models) for predicting outages (note the message behaviors do not include information specifying any deficient grid element).
  • the confidence level can be increased and the models re- selected or the results further filtered to match the repair crew capacity. If the models predict fewer outages than the repair crew can handle, then the confidence level can be decreased and the models re- selected or the results less filtered to match the available capacity of the repair crew. For example, the confidence level can be decreased prior to weekend days, seasonal vacation periods or other times when the repair crew has less capacity for repairs so that required repairs may be minimized during those periods when the repair crew has less capacity. [0093] Referring to node 1509, the decision tree shows that out of the total 10100 locations or premises analyzed, the 12 premises that generated both NIC.PD.C.U and Zero V.R.
  • event messages have a probability of 85.714% with respect to reporting a failure.
  • the decision tree shows that out of the total 10100 locations or premises analyzed, the 5 premises that generated both NIC.PD.C.U and Last Gasp event messages have a probability of 100% with respect to reporting a failure.
  • the decision tree shows that out of the total 10100 locations or premises analyzed, the 5 premises that generated both Zero V.R. and P.F.C. event messages have a probability of 100% with respect to reporting a failure.
  • the decision tree shows that out of the total 10100 locations or premises analyzed, the 9 premises that generated NIC.PD event messages at a concentration greater than 0.2% have a probability of 90% with respect to reporting a failure.
  • the decision tree may then be used to identify types of events that may be correlated in order to ascertain whether a single premises will experience a power failure.
  • end nodes may also be used to validate an absence of an outage. For example, if a message behavior corresponding to node 1539 was determined and yet an outage report was received from associated a suspect premises, then it may be determined that the outage report is erroneous, particularly if the outage report is from a premises generating chronic outage reports. Thus a repair crew would not be dispatched and a "bad roll" would be avoided. In a further example the approach could be applied to a group of premises to determine an absence of outages in a multiplicity of premises within the group of premises.
  • FIG. 8 through FIG. 15 illustrate among other things, a model generator coupled to the message database for generating a multiplicity of models based upon the messages and the outage reports, each model identifying a defective grid element based upon a message behavior of a smart meter associated with the defective grid element, the message behavior including a plurality of messages of a plurality of message types received from the smart meter, and a model selector for selecting a plurality of models from the multiplicity of models based upon a confidence factor.
  • a model generator coupled to the message database for generating a multiplicity of models based upon the messages and the outage reports, each model identifying a defective grid element based upon a message behavior of a smart meter associated with the defective grid element, the message behavior including a plurality of messages of a plurality of message types received from the smart meter, and a model selector for selecting a plurality of models from the multiplicity of models based upon a confidence factor.
  • the present disclosure has numerous other applications; it may be used to predict failures on other types of devices (single phase laterals or transformers) or in connection with other types of services (neutral legs on a service).

Abstract

A system and method is disclosed for using AMI smart meter messaging types and data mining decision trees to determine if local equipment failure is present. The system and method may be used to predict impending failure based upon smart meter message behaviors and to create proactive investigation tickets. The predictions models may be generated from a big database of smart meter messaging and customer outage reports. The system and method can be applied to detect failures of higher level device equipment and may be incorporated into customer service processes. The system and method may also be used to determine customer owned equipment failures for referral to electricians.

Description

OUTAGE PREVENTION IN AN ELECTRIC POWER DISTRIBUTION GRID USING
SMART METER MESSAGING
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application claims the benefit of United States Non-Provisional Patent Application Serial No. 15/002,180 filed January 20, 2016 and United States Provisional Patent Application Serial No. 62/173,039, filed June 9, 2015, the entire contents of which are herein incorporated by reference.
FIELD OF THE DISCLOSURE
[0002] The present disclosure relates to a system and method for utilizing smart meter messages to predict deficient elements of an electric power distribution grid prior to a customer reporting of an outage.
BACKGROUND
[0003] Electrical power distribution grids can be implemented as radial, loop or network type systems. The distribution grids are arranged and interconnected to a substation in different ways depending on the type of system configuration. However for each type of distribution system configuration, the distribution circuits (commonly referred to as feeders and lateral feeders) distribute power delivered from the substation to loads at premises coupled to the grid through smart meters.
[0004] Various types of faults can occur in an electrical distribution system, some of which result in power outages, i.e., the loss of electric power service to customers. For example, a short circuit fault causes the protective element upstream of the fault to open isolating the short circuit fault from the grid. A short circuit may be caused by a tree branch contacting power lines during a storm for example. Customers downstream of the opened protective element become de- energized resulting in an outage. Another type of fault is an open conductor element fault that similarly causes the downstream customers to experience a power outage. An open conductor element may be caused by a power line snapping during a storm, or a coupling joining two power lines becoming deficient and then failing thereby resulting in the open conductor.
[0005] Power outage analysis conventionally relies on customer phone calls made to the utility company in the event of a power outage as the main information source for such analysis. This process can be quite slow because many customers may not call to report an outage, and those who do report an outage may wait a relatively long period of time to report the outage, often assuming a neighboring customer will call in their stead. With the wide deployment of AMR (automatic meter reading) and AMI (advanced metering infrastructure) technologies in power distribution systems, timelier message information is available from smart meters. However, smart meters may not be designed to specifically detect deficient grid element and generate a corresponding message. Furthermore smart meters generate a multiplicity of message types, each of which has been shown to be an unreliable indicator of a service interruption that would qualify as an outage that requires a repair to an element of the power distribution grid.
[0006] When a customer contacts the utility to report an outage, a repair crew is then dispatched to determine the deficient grid element, performing fault reparation and ultimately service restoration. Without the benefit of an AMI, the repair crew may not be able to be certain if the outage is a single premises outage or an outage to multiple premises. However, repairs have not been implemented until after a customer experiences an outage and has taken the steps to contact the utility. Then the customer waits without power until a repair crew is dispatched, the cause of the outage is determined and power restored, thereby leading to a degraded quality of service and increasing customer dissatisfaction. SUMMARY
[0007] The following presents a simplified summary of the disclosure in order to provide a basic understanding of some aspects of the disclosure. This summary is not an extensive overview of the disclosure. It is intended to neither identify key or critical elements of the disclosure nor delineate the scope of the system and method disclosed herein. Its sole purpose is to present some concepts of the disclosure in a simplified form as a prelude to the more detailed description that is presented later.
[0008] The system and method of the present disclosure generally relate to the use of AMI messaging types and data mining decision trees to determine if local equipment failure is present. The system and method may be used to predict impending failure and to create proactive investigation tickets. The system and method can be applied to detect failures of higher level device equipment and may be incorporated into customer service processes (troubleshooting during first customer contact). Further, the system and method may also be used to determine customer owned equipment failures for referral to electricians (inside the premises trouble).
[0009] The following description and the annexed drawings set forth in detail certain illustrative aspects of the disclosure. These aspects are indicative, however, of but a few of the various ways in which the principles of the system and method disclosed herein may be employed and the system and method disclosed herein is intended to include all such aspects and their equivalents. Other advantages and novel features of the system and method disclosed herein will become apparent from the following detailed description of the system and method disclosed herein when considered in conjunction with the drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0010] FIG. 1 illustrates one implementation of a method to create a Single premises No Current (SNC) ticket based on a predictive algorithm.
[0011] FIG. 2 illustrates one implementation of a method to investigate single no current prediction tickets created through the process illustrated in FIG. 1.
[0012] FIG. 3 illustrates the monitoring of AMI smart meter messages to ascertain the likelihood of an SNC outage in accordance with one implementation.
[0013] FIG. 4 illustrates a graphical user interface for the premises troubleshooting tool in accordance with one implementation.
[0014] FIG. 5 illustrates a chart showing last gasp messages observed in order to ascertain whether a true sustained power outage has taken place in accordance with one implementation.
[0015] FIG. 6 illustrates a chart that reflects intermittent power conditions in accordance with one implementation.
[0016] FIG. 7 illustrates at a high level the flow of information from a smart meter in accordance with one implementation.
[0017] FIG. 8 illustrates a methodology for selecting power failure prediction models in accordance with one implementation.
[0018] FIG. 9 illustrates a user interface that may be used to display notifications alerting a utility that a power failure is likely to occur in accordance with one implementation.
[0019] FIG. 10 illustrates an exemplary set of event messages retrieved from an AMI database in accordance with one implementation.
[0020] FIG. 11 illustrates a combination of ticket and interruption types that may be analyzed in one implementation. [0021] FIG. 12 illustrates an exemplary histogram or time series plot of AMI event messages in accordance with one implementation.
[0022] FIG. 13 illustrates a sample of the data prepared for input into a data mining engine in accordance with one implementation.
[0023] FIG. 14 illustrates a sample screenshot of a data extraction process in accordance with one implementation.
[0024] FIGs. 15A-C illustrate a decision tree that may be used to predict power failures in accordance with one implementation.
[0025] FIG. 16 is a simplified block diagram illustrating components of a metered electric utility distribution system.
DESCRIPTION OF ILLUSTRATIVE IMPLEMENTATIONS
[0026] The system and method disclosed herein will now be described more fully hereinafter with reference to the accompanying drawings, in which preferred implementations of the system and method disclosed herein are shown. The system and method disclosed herein may, however, be implemented in many different forms and should not be construed as limited to the implementations set forth herein. Rather, these implementations are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the system and method disclosed herein to those skilled in the art.
[0027] The systems and methods described herein may be implemented as a method, data processing system, or computer program product. Accordingly, the description may take the form of an entirely hardware embodiment, an entirely software embodiment, an embodiment combining software and hardware aspects, a computer program product on a computer-usable storage medium having computer readable program code on the medium, a non-transitory computer readable storage medium, or combinations thereof. Any suitable computer readable medium may be utilized including, but not limited to, static and dynamic storage devices, hard disks, optical storage devices, and magnetic storage devices.
[0028] The present system and method is described below with reference to illustrations of methods, systems, and computer program products according to the disclosed implementations. It will be understood that blocks of the illustrations, and combinations of blocks in the illustrations, can be implemented by computer program instructions, hardware devices, or a combination of both. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a particular machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, implement the functions specified in the block or blocks.
[0029] The disclosed system and method may be implemented on one or more computing devices, including one or more servers, one or more client terminals, including computer terminals, a combination thereof, or on any of the myriad of computing devices currently known in the art, including without limitation, personal computers, laptops, notebooks, tablet computers, touch pads (such as the Apple iPad, SmartPad Android tablet, etc.), multi-touch devices, smart phones, personal digital assistants, other multi-function devices, stand-alone kiosks, etc.
[0030] First, turning to FIG. 16, FIG 16 illustrates a simplified block diagram of elements of an electric power distribution system. The distribution system 100 includes a power generation source 110 that is operative to generate electric power in accordance with one or more technologies and to distribute the electric power via transmission lines 120. In that regard, source 110 may employ or included a hydro-electric plant, a fossil fuel or coal-fired plant, a nuclear plant, a wind farm or a solar array or collector and attendant structures, a geothermal plant, or a combination of these and other facilities that are known or developed in accordance with known principles and technologies to generate electric power. Nuclear reactors, steam, gas, or water- driven turbines, rotors, stators, solar panels, cooling towers, and various other ancillary equipment and machinery are omitted from FIG. 16 for clarity, but may be employed at or in cooperation with source 110 as is generally known, and it will be appreciated that these omitted details may be application-specific. In some implementations, transmission lines 120 may carry high- voltage Alternating Current (or "AC") electric power from source 110 to a desired destination as is generally known, though other technologies for distributing electric power across distances, as may be appropriate for a particular application and based upon technological readiness and suitability, are contemplated. In particular, the present disclosure is not intended to be limited by the structural or operational characteristics of source 110 or transmission lines 120, or by the nature of the electric power and the manner in which it is generated or transmitted.
[0031] In some practical applications, electric power generated at source 110 may be transmitted via transmission lines 120 to a power transfer substation 130 which, in turn, distributes electric power to transformers (Txi, Tx2, . . . Txn) 140-141 for subsequent transmission to a multiplicity of premises or a group of premises 144-145 having a corresponding group of smart meters that are generally represented as meters (Mi, M2, . . . Mn) 150-151. It is noted that electric power distribution system 100 may generally include more power generation sources (such as source 110) and more substations (such as substation 130) than depicted in FIG. 16, and that each such substation 130 may serve many more transformers (such as transformers 140) than illustrated. Similarly, each such transformer 140, such as specific transformer 141, may provide electric power to more premises and smart meters (such as meters 150-151) than are depicted in FIG. 16. In that regard, the present disclosure is not intended to be limited by the simplified topography represented in the drawing figures.
[0032] In operation, substation 130 may modify or condition the electricity received from source 110 such that it may then be transmitted to transformers 140. For example, it may be desirable in some circumstance to step down (or to step up) voltage via one or more substation transformers 131, or to phase-shift or otherwise to adjust current phase or amplitude, for instance, to achieve a desired power function as specified by the kind of load and/or to minimize energy lost in the distribution system. Various techniques are known or may be developed to condition electric power at substation 130, and the present disclosure is not intended to be limited by the operation of substation 130 or by any technical procedures executed or functionality employed there.
[0033] Similarly, transformers 140-141 may be configured and operative further to condition the electric power received from substation 130 such that it is suitable for delivery to customers or loads associated with smart meters 150-151. Voltage manipulation, current manipulation, or a both, may be employed in various situations; the nature and extent of such conditioning may be dependent upon the specifications and operational characteristics of the smart meters 150-151 to which electric power is delivered, for example, or upon governmental regulations, technological or infrastructure capabilities, or a combination of these and other factors. The present disclosure is not intended to be limited by the technologies employed at transformers 140-141.
[0034] Smart meters 150-151 may be configured to measure electric power consumed at each corresponding premises and generate a multiplicity of message types related to meter electric power usage at a particular premises such as a residence, building, business location, or some other specific site. In some circumstances (where available from and supported by a particular utility service provider, for instance, or where required by a state or local utility oversight commission or committee), meters 150-151 may be operated in accordance with Advanced Metering Infrastructure (or "AMI") specifications or protocols.
[0035] In that regard, meters 150-151 may be configured as or include what are generally referred to as "smart meters" or "AMI meters." In operation, when implemented as a smart meter, meters 150-151 may monitor, sense, record, or otherwise track electrical consumption (or "usage") digitally and at predetermined or dynamically adjustable intervals; subsequently or effectively contemporaneously, meters 150-151 may transmit usage data (e.g., using a secure radio frequency band or other telecommunications methodology) to other devices installed on or otherwise communicably coupled to the electrical grid or to a communications network (such as AMI network 160) with which meters 150-151 may exchange data. In this context, it will be understood that "usage data" may refer to raw data (i.e., unmodified or unprocessed data as they are collected and that represent the amount of electrical energy consumed at the location metered by meters 150-151) or to data and other information relating to or derived from such raw data (e.g., readings or data points that may be time-stamped or otherwise processed to provide more information than simply an aggregate or a measure of cumulative consumption). For example, in some instances, an aggregate usage value may be transmitted from such a smart meter, while in other instances, detailed, time-dependent usage rates may be transmitted; the specific type and amount of data collected and processed at, and transmitted from, meters 150-151 may be application-specific and may vary in accordance with processing or computational capabilities of hardware components deployed in, as well as software functionalities implemented at or in cooperation with, meters 150-151. These factors may be affected by technological or economic considerations, for example, or may be dictated or influenced by applicable governing bodies or governmental regulations.
[0036] As illustrated in FIG. 16, smart meters, such as meters 150-151, may be communicably coupled to AMI network 160 such that messages including usage data collected (and possibly processed) by meters 150-151 may be transmitted to AMI network 160.
[0037] In operation, AMI network 160 may be communicably coupled to a utility provider's facilities and computer systems (hereinafter generally referred to as "systems" 180). The AMI network may include the Internet, for instance, or any other packet-switched data network, virtual private network, proprietary or public wide area network ("WAN"), or any other communications network capable of bi-directional data communication between AMI network 160 and systems 180.
[0038] At a service provider's facilities, systems 180 generally include processing and billing systems, various monitoring, customer service, troubleshooting, maintenance, load balancing, accounting, and other types of activities that may be used to operate a utility service. In some instances where such activities are computationally expensive or require a great deal of processor power or communication bandwidth, some implementations of processing may be distributed across many processing and memory storage resources, and even distributed across buildings as is generally known in the art. Accordingly, though systems 180 are depicted in FIG. 16 as a single, unified entity, it will be appreciated that block 180 may represent any number of physical buildings, computer server facilities, and other physical and logical resources owned, operated, or otherwise under the control of the relevant utility service provider.
[0039] The discussion below turns to more particular implementations of the present disclosure. A customer such as a customer located at premises 1, 144 may experience an interruption in service, or an outage. An outage may result from an electric power distribution grid having any of a multiplicity of elements becoming deficient of failing. The elements include any component of the grid coupling power generation source 110 to premises 144, including power generation source 110. If an element between or including power generation source 110 and transformer 140 becomes deficient and fails, then a multiple premises outage may result.
[0040] A single premises outage may occur if an element between transformer 151 and premises 144 becomes deficient or fails. A single premises outage may occur for any of a multitude of reasons. In one example transformer 142 has a power coupling 250 for providing power to premises 1 through n, 144-145. If a portion of power coupling element 250 becomes deficient, such connection corrosion or insulation deterioration, then an outage at premises 144 may occur. If power distribution line 252 becomes deficient, either by corrosion or high impedance leakage or a fault in an underground conduit, then an outage at premises 144 may occur. If a deficiency in power coupling elements 245, which may couple copper lines 256 from premises 144 to aluminum lines 252 results is a failure, then an outage at premises 144 may occur. If copper power distribution line 254 becomes deficient, either by corrosion or high impedance leakage or a fault in an underground conduit, then an outage at premises 144 may occur. If meter Ml, 150, becomes deficient, then an outage at premises 144 may occur.
[0041] In response to an outage, the customer contacts the utility 180 via telephone 258 and an outage report or outage notice is generated by outage processing process 260. Telephone 258 is coupled to the utility via a public switched telephone network (PSTN) 190. Is should be noted that the PSTN network 190 for communicating telephone calls from a customer at premises 144 to utility 180, is a separate network from the AMI network 160 used for communicating smart meter messages from smart meter 150 at premises 144 to utility 180, even though various components may be shared between networks, such as internet or cellular networks. The PSTN network 190 is used for customer communications with the utility and the AMI network 160 is used for smart meter message communication with the utility. The PSTN and AMI networks operate independently and facilitate different types of communication. In other examples, alternate communication devices, other than phone 258, may be utilized to communicate an outage from a customer to the utility while remaining within the scope of this description. Other devices may include smart phones, personal computers or other devices for customers to communicate with the utility.
[0042] In response to an outage report, outage processor 260 process the outage report and determines an appropriate response. Ticket generator 270 then generates a ticket via a ticket output device in the form of a humanly observable repair ticket (such as a printed page or display on a computer screen) indicative of the prediction, thereby enabling a repair crew 275 to repair the deficient grid element associated with the outage at the premises. The findings of the repair crew may then add to the outage report. The outage processor may use a number of approaches known to those familiar in the art for determining an appropriate response. Such approaches may include analyzing a multitude of outage reports from a multitude of premises. In order to keep customer satisfaction high and to reduce maintenance costs, it is desirable to reduce the amount of tickets and the amount of repair work associated with the grid.
[0043] Ticket type represents what the customer reported via the phone system and before performing a field investigation include:
SUB = substation,
FDR = feeder,
LAT(U)= Lateral (overhead or underground), TX(U) = transformer (overhead or underground),
SNC(U)= single no current (overhead or underground), used to report PNC conditions, and
NLS= no loss of service, used to report power quality or other field conditions which cannot be deferred to regular work (reachable low wire, for example).
[0044] Interruption type represents the category of the outage after the field investigation and repairs are complete and include:
SUB - substation caused failure,
FDR - feeder caused failure,
LAT - lateral caused failure,
TX - tx caused failure,
OTH-SEC - multiple customer service failure (including failures of handholes or other junctions downstream of a distribution tx),
OTH-SV - single customer service drop failure,
OTH-MTR - meter caused single customer outage, and
NLS - no power loss observed.
[0045] Tickets issued to correct problems with feeders (FDRs), which may be defined as main power lines traveling out of a substation, with laterals (LATs), which may be defined as fused power lines traveling off the main line feeder, with substations (SUBs), and with transformers (TXs) are conventionally responsible for the ample majority of service interruptions. FDR, LAT, SUB and TX tickets may be responsible for more than 97% of customer outages but only produce less than 12% of the repair ticket volume.
[0046] By contrast, single premises events may be responsible for less than 3% of the customer outages while generate over 74% of the repair ticket volume, thereby making up the majority of restoration work of an electric power distribution grid utility's maintenance fleet. Single premises events include single no current (SNC) tickets, which may relate to a single customer or meter reporting a loss of power, no loss of service (NLS) tickets, which may relate to an outage ticket type generated when a customer is reporting that there is no sustained outage or interruption of electric service at the time, and location (LOC) tickets, which may relate to an outage ticket type where the suspected device cannot be identified. Thus, single premises events result in high number of tickets. Reducing single premises events may significantly improve the quality of service provided to customers of a utility providing electric power to customers. Consequently, it may be advantageous to predict the occurrence of single premises events before an outage is reported.
[0047] In the present disclosure, a "bad truck roll" may be defined as a situation when either no utility problem exists or an investigator is unable to determine the problem and yet the investigator makes a trip to a premises or location to conduct an investigation based upon a ticket. One potential advantage of the present disclosure is to reduce investigator volume.
[0048] A single premises event or outage may typically result from a failure related to grid elements 250-256. Smart meter 150 is able to generate a message as a result of an outage, such as a last-gasp message. As will be detailed with respect to FIG. 6, reception of a single last gasp message (or multiple last gasp messages generated by a single smart meter) by AMI processing function 189 is typically not reliable enough to substitute for an outage report generated in response to a customer call. Furthermore, elements of the power grid, such as elements 250-256 become deficient slowly before a failure results in an outage. Power couplings may slowly corrode before failure, insulators may slowly become contaminated before failure, underground cable may slowly degrade before failure, vermin may slowly destroy elements before failure. In order to reduce the occurrence of single premises events and improve the quality of power delivered to the customer, it may be advantageous to detect elements becoming deficient before a failure resulting in an outage occurs.
[0049] While smart meters generate a multiplicity of messages, the messages do not include information specifying a deficient grid element. For example, smart meter messages do not specify deficient power couples 250 or 254, nor do they specify deficient lines 252 or 256. However, the inventors have discovered how to determine a smart meter message behavior that is indicative of deficient grid elements and predictive of future outages, and then apply the smart meter message behavior to messages received by smart meters in the grid to identify a suspect premises and generate a ticket to enable a repair crew to begin investigation and/or repair a deficient grid element prior to an outage being reported by a customer.
[0050] While the objective of predicting and repairing a grid element prior to an occurrence of a single premises outage task is simply stated, realizing the objective compounded by the enormity of data and the extremely low percentage of premises experiencing single premises outages. For example, in a twenty four hour period, in a group of six hundred thousand premises, corresponding smart meters may generate over one million two hundred thousand event messages having over two hundred and eighty message types. In that twenty four hour period, only sixty single premises outages may occur, or one in ten thousand or zero point zero zero one percent. Finding message patterns for predicting single premises outages in such a mass of big data when the smart meter messages do not themselves specify a deficient grid element has been compared to finding a needle in a haystack.
[0051] An objective of the present disclosure is to generate tickets before receiving a customer call to report an outage. Still another objective of the present disclosure is to restore power at a premises before a customer residing in that premises calls to report an outage. Still another objective of the present disclosure is to dispatch a repair crew before a customer residing in that premises calls to report an outage. While repairing a deficient grid element prior to an outage results in a high quality of customer service, repairing deficient grid element resulting in an outage before a customer calls and an outage report is generated is an improvement in customer service quality and customer satisfaction. Further still, dispatching a repair crew to a premises prior to receiving an outage call is also an improvement in customer service, even though the customer had to call the utility, the utility can respond that a deficient element was detected and a repair crew has already been dispatched prior to the customer placing the call.
[0052] Thus, FIG. 16 shows An electric power distribution system 100 comprising an electric power distribution grid having a multiplicity of elements 250-256 capable of becoming deficient, a multiplicity of smart meters 150-151 coupling the electric power distribution grid and a corresponding multiplicity of premises 144-145, each smart meter measuring electric power consumed at each corresponding premises and generating a multiplicity message types related thereto, the multiplicity of message types not including information specifying a deficient grid element. FIG. 16 also shows an automated meter infrastructure network 160 coupled to the smart meters for communicating the messages from the smart meters and a message database within AMI processor 189 coupled to the automated meter infrastructure network for receiving messages from the multiplicity of smart meters, and an outage processor 260 for receiving outage reports through a communication network 190 different from the automated meter infrastructure network 160, each outage report associated with a failed deficient grid element. FIG. 16 also shows an outage predictor 265 for processing the additional messages with the plurality of models to predict a likelihood of an outage at a suspect premises based upon messages from smart meters, and a ticket generator 270 for generating a ticket enabling a repair crew to repair a deficient grid element associated with the suspect premises.
[0053] While a detailed description of a process for identifying message behaviors related to deficient smart grid elements is included with respect to FIG. 4 through FIG. 15, introductory comments regarding the process for identifying message behaviors and affecting repairs are provided in the following paragraphs and with respect to FIG. 1- FIG. 3.
[0054] In accordance with one implementation, a data mining process may be applied to messages having event data transmitted by smart meters. A data mining process may be defined as a process of analyzing data from many different perspectives, which may allow, for example, viewing information from many different perspectives or "angles" at the same time. For example, instead of sampling data points, the entire population of available data may be used to find patterns predictive of outages. In accordance with one implementation, the process data mines smart meter messages to look for patterns indicating whether a future single premises interruption may be likely to occur.
[0055] To predict outages, one of the methods of the present implementation may rely on event correlation. Meters send structured messages, which include event information, over the AMI network. A history of smart meter messages received over the AMI network and a corresponding history of outage reports received over the PSTN network may be analyzed. In one implementation, a process receives these structured smart meter messages and filters out duplicate messages. Duplicate messages may be received since the same message generated at a meter may be routed through different nodes in a mesh based AMI network. After scrubbing the data and filtering out duplicates, the messages and outage report database may be organized into a data structure. The data structure associated with the meter messages may relate to many attributes of the meter. For example, an AMI meter may generate over 280 types of event messages.
[0056] Once the data has been scrubbed, it may be processed by a data mining engine. In one implementation, the data mining engine may be IBM's SPSS tool (http://www- 03.ibm.com/software/products/en/spss-decision-trees). A data mining engine may be implemented as statistical analytical software used for data analysis, data mining and/or forecasting. Use of the data mining engine may enable users to import (from a variety of sources including databases, flat files, IBM Cognos, Excel, XML, etc.), manage, and analyze data in a graphical user interface. It may also allow the user to construct graphs and charts, tables, cross- tabulations, build forecasting models, decision-trees, and perform a number of advanced statistical functions and export them to the desired format.
[0057] The data mining engine may generate equations describing message behaviors based on any of a number or theories including for example, a natural theory of entropy. These equations or algorithms may then be used to predict outages. In one embodiment, the equations are represented as a visual and textual breakdown of a decision tree. The decision tree is selected based on a comparison of different correlations among meter events and selecting the decision tree which best predicts failures. The decision tree analysis may be used to model a series of events and look at how they affect an outcome by calculating a set of conditional probabilities based on different scenarios.
[0058] In one implementation, a CHAID (CHi-squared Automatic Interaction Detection) decision tree may be selected for failure prediction. In another embodiment, a different process may be selected for splitting and pruning a decision tree, and noise handling may be selected, such as C5.1 or other decision tree process known to those skilled in the art. Once one of the models or trees is selected, algorithm logic may be defined and delivered. The algorithms are capable of generating prediction models for predicting single premises outages before they occur based upon messages from smart meters. The final algorithms look for groups of event messages in a hierarchical method. In one implementation, application of the decision trees shows that premises with SNC (interruption type SEC, SVC) have smart meters that send certain types of messages or message behaviors at a rate or occurrence or in certain concentrations 48-24 hours before an outage occurs.
[0059] In one implementation, the process may generate a NLS ticket, then upgrade to a SNC upon receiving customer call reporting an outage. In addition to predicting single customer outages, the process can help reduce the number of "bad truck rolls."
[0060] Once message behaviors able to predict deficient elements of the grid have been determined, messages from smart meters may be analyzed to predict outages. FIG. 1 and FIG. 2 describe a use of the message behaviors for implementing corrective actions.
[0061] FIG. 1 illustrates one implementation of a method to create a SNC ticket based on a predictive algorithm using smart meter message behavior. In step 101 the process opens a list of flagged premises. The premises may be flagged by an outage predictor which flags premises based on at least an output smart meter messages processed by any of a plurality of decision tree based models. In step 103 the process filters for premises flagged in last 24 hours. In optional step 105 the process filters premises for current area of interest. In step 107 the process determines whether there are any new premises flagged, and if no new premises are flagged, then in step 109 the process closes a list of flagged premises. The process may optionally go on to notify a team, such as an evaluation team, of the results. If new premises are flagged, then in step 111 the process determines whether a premises had a ticket in the last 48 hours. If the premises had a ticket in the last 48 hours, then in step 113 the process monitors the premises for another 48 hours, and then the process in step 115 determines whether the premises is continuing to be flagged, and if the premises does not continue to be flagged, the process ends.
[0062] In step 117, the process opens a premises troubleshooting tool (see Fig. 4) after it has been determined that a premises did not have a ticket in an initial period of 48 hours (step 111) or after it has been determined that a premises continues to be flagged after being monitored for an additional 48 hour period following the initial 48 hour period (steps 113 and 115). The premises troubleshooting tool allows selection of a date range and assignment of a premises number in step 119. Step 121 records premises information; step 123 records daily average kwh usage; step 125 records meter ping result (e.g., ping for instantaneous voltage data); step 127 records event summary details; and step 129 opens a ticket summary tab.
[0063] In step 131 the process determines if a new NLS ticket is required, and if no new NLS ticket is required the process ends. If a new NLS ticket is required, then the process, in step 133, opens TCMS and creates a new trouble call management system ("TCMS") ticket including the NLS ticket. In step 135 the process adds inspection instructions and note "Technology" to ticket thereby indicating that the ticket is created by the utility predictive model and not a conventional customer outage report. In step 137 the process records details and comments on the flagged premises list. In optional step 139 the process notifies a team, such as the pilot team, of the results.
[0064] FIG. 2 illustrates one implementation of a method to investigate single no current prediction tickets created through the process illustrated in FIG. 1. In step 201 the process receives an NLS prediction ticket generated in FIG. 1. In step 203 the process investigates the ticket as shown on the restoration process for restoring power (which is part of a standard repository of business processes: RST— 0.6). In optional step 205, the process generates a signal for notifying the customer of proactive equipment failure evaluation details. In step 207 the ticket is sent to a repair crew with instructions to follow for the repair. The ticket includes extra instructions from the trouble call management system ("TCMS"). The extra instructions may be included in the customer remarks field. The TCMS extra instructions may be used to enhance the management of outages or identification of potentially defective grid elements. For example, a TCMS ticket may be sent to a crew with specific steps that let crew know what to do in the event of intermittent power problems such as conduct additional due diligence, follow through extra steps and document in the system. In step 209 the process determines whether there is a deficient grid element resulting in equipment failure present based upon the finding or investigation of the crew. If there is no equipment failure the process determines whether the ticket needs to be referred for equipment repair (step 215). If the ticket does not need to be referred, the ticket is closed (step 217). If there is equipment failure (step 209), then as part of the process of the investigation of the crew, a photograph may be taken (step 211) of the equipment failure, and is attached to the ticket, and in optional step 213 the customer of a proactive equipment failure investigation result is notified. [0065] If a ticket indicating a deficient grid element needs to be referred for repair (step 215), then in step 219 the deficient grid element equipment is repaired. The equipment may be repaired in accordance with a standard business process which may be described for example in restoration process RST-0.6. In step 221 the process determines whether the damaged equipment was removed during repair. If the damaged equipment was not removed, then the process in step 223 updates the ticket as shown on restoration process RST— 0.6 and the ticket is closed. If the damaged equipment was removed, then in step 225 the process quarantines defective equipment. The quarantined equipment may be used by an evaluation team to further enhance the failure prediction model. For example, damaged cables may be removed and held on to so that the utility can see the damage and add the results to the prediction model generator. The process updates (step 227) the ticket as shown in restoration process RST 0.6, and the ticket is then closed (step 237).
[0066] FIG.3 illustrates a decision tree based model for the monitoring of AMI smart meter messages to ascertain the likelihood of an SNC outage in accordance with one implementation. It has been discovered that premises with SNC outages exhibit a message behavior that results in the smart meter sending certain types of messages in certain concentrations 24-48 hours before the SNC outage. In the illustrated embodiment, if (as shown by Node 0) a premises sent one or more "NIC Based Power Fail Detect Disabled" (NIC or Network Interface Card may be defined as the computer chip in the smart meters that enable messaging and communications) and (as shown by Node 28 and Node 29) less than 1.5% of its total messages sent were "Service Error Cleared" (which is a type of AMI event indicating that there was an issue detected by the meter but that it has cleared itself), then using the predictive algorithm the process determines that there is more than 95% chance that the premises associated with the smart meter will have an SNC interruption in the next 24 hours. Accordingly, the premises will be flagged and processed in accordance with the process of FIG. 1.
[0067] FIG. 4 illustrates a graphical user interface ("GUI") produced by the premises troubleshooting tool in accordance with one implementation. This tool may be used for investigating a flagged premises. The GUI 413 prompts the user for premises information, which is then uploaded into the tool. The GUI displays detailed premises information 401 for a premises that has been flagged by the process of the present disclosure, Event Summary information 403, detailed Event information 405, and ticket information 407. The GUI also displays a graph plotting data corresponding to power usage in kWh and corresponding to measured voltage in volts root mean square (Vrms). In the illustrated exemplary implementation, a voltage and usage anomaly 409 caused the premises to be flagged. The anomaly shows the voltage and power dropping to zero after September 2. The GUI also displays summary voltage and usage information 411.
[0068] FIG. 5 illustrates a chart 600 showing last gasp messages which do not represent a true sustained outage. In one implementation, a last gasp message may be defined as an event generated when an AMI meter detects a zero voltage event for a period of time (e.g., 45 seconds). The voltage and power at a premises are shown dropping to zero after June 11. After the drop numerous smart meter messages including last gasp messages are generated as represented by dots occurring after June 11. While focusing on the valuable messaging from the AMI network chatter may yield important information about outages, because the amount of last gasp messages illustrated in FIG. 5 does not represent a true sustained outage, the last gasp messages alone are not entirely accurate for generating reliable single premises tickets. [0069] FIG. 6 illustrates a chart 700 that reflects intermittent power conditions. In one embodiment, the processes for predicting power failure exploit smart meter message patterns with intermittent or partial power problems.
[0070] FIG. 7 illustrates at a high level the flow of information (notifications, voltage reads, event messages, etc.) from a meter. In addition to outage notifications generated by a smart meter, each smart meter generates a significant number of voltage read messages and significantly more event messages. In an example power distribution system having six hundred thousand smarter meters, over one point two million event messages across over two hundred and eighty categories or message types (none of which may specify a deficient grid element) may be generated in a twenty four hour period while only sixty single premises outages may occur within the period. Thus processing one point two million messages per day to predict sixty outages before a customer generate outage report occurs may be akin to finding a preverbal needle in a haystack. Furthermore, a typical utility may have very many more smart meters in the power distribution system, over four million smart meters for example. The resulting very large amount of data may be stored in a database. The database may be a database adapted for storing big data files in a big database. Such a database may be an extensible relational database platform using a massively parallel processing approach such as a Greenplum database, as is known to those familiar with the art. Historical data may be housed in the database and then analyzed using data mining techniques in accordance with the present disclosure. These notifications, reads, and event messages may include last gasp messages, power failure count messages, NIC power down messages, inversion count notifications, zero voltage messages, DC detection cleared messages, and other types of messages as would be recognized by those skilled in the art. [0071] FIG. 8 illustrates a methodology for generating power failure prediction models in accordance with one implementation of the disclosure. Numeral 801 represents the database that contains meter event data. Numeral 805 represents a step of scrubbing meter event data 803 collected over a two day period. The data is scrubbed for duplicates, errors, etc. Once the data is pre-processed, it is loaded to a data mining engine 807. The data mining engine considers meter event message information as well as customer generated power outage information. As a result of the data mining, different models may be created 809 for the prediction of power failures. In accordance with one implementation, two decision tree models, such as CHAID and C5.1 are selected such that event data can be fed into the algorithms and power outages can be predicted before being reported by a customer. These algorithms may be represented by decision trees that take into account several types of messages and their concentration. The Prediction Algorithm rectangle of FIG. 8 shows an example of an algorithm indicative of a message behavior based upon a branch of a decision tree model, e.g., { message 1+ (message 2 < X% concentration) OR (message 2 > X %+ message 3 < Y concentration)) ( message 4 < Z % concentration) }.
[0072] FIG. 9 illustrates a user interface that may be used to display notifications 901 alerting the utility of a prediction that a power failure is likely to occur at a suspect premises. The user interface may display premises information 909, voltage readings 907, meter event summary 903, and a history of events 905. The aforementioned decision trees are applied to the event data in order to ascertain whether to display an alert 901. The information displayed in FIG. 9 including notification 901 may allow a ticket generator to generate a ticket enabling a repair crew to repair a deficient grid element associated with the suspect premises. Given the predictive nature of the notification the repair may occur prior to a failure of a defective grid element, or prior to a generation of an outage report. [0073] FIGs. 10-15 describe more specific implementations for predicting equipment failure in an AMI network. FIG. 10 illustrates an exemplary set of event messages retrieved from an AMI database (e.g., a GreenPlum database). In one implementation, event messages from a time period may be selected for extraction. Altogether, more than 100 million event messages may be extracted and saved in a CSV file. In the exemplary extraction process illustrated at least partially in the SQL Editor window of FIG. 10, PostGreSQL is one of the possible tools known to those familiar with the art that may be used to access a GreenPlum database.
[0074] Although only a partial list of messages is shown in FIG. 10, the following AMI meter event data or information may be extracted from GreenPlum:
Premises Number
Event Name
Event Timestamp
Transformer Number
Feeder Number
[0075] In addition to event messages, voltage readings of zero (0V) may be classified as notable events and so recorded. With the messages extracted, a complex data cleanup may then be started. The data cleanup may include defining and removing duplicate messages and messages from commercial industrial meters or other meters not associated with a single premises. A duplicate message may be defined as any message which has been sent more than once from the same meter within a time period, such as one second. When the data scrubbing is completed, only discrete events from residential meters remain in accordance with one implementation.
[0076] After the event data extracted is ready for analysis, the process may query all single premises trouble or outage tickets from an outage processor which may be included in a Power Delivery Distribution data warehouse. Premises which experienced an outage during the time frame of interest may be noted, and a column added from the outage processor to the corresponding data record to indicate whether an event message was sent by a meter before (which may be defined as any event which occurred more than 6 hours before the meter was interrupted), during (which may be defined as any event from 6 hours before the meter was interrupted until power to the meter was restored), or after (which may be defined as any event which occurred after power was restored to the meter) an outage event.
[0077] FIG. 11 illustrates a combination of ticket and interruption types that may be analyzed in one implementation. FIG. 11 shows eight columns with corresponding ticket types and seven rows with corresponding interruption types as analyzed by a repair crew in response to a ticket. In one implementation, tickets which started as a single premises interruption (single no current - SNC/single no current underground - SNCU) and ended as a secondary (SEC) or service (SV) interruption (e.g., service indication of intermittent problems) are considered as part of the algorithm development for predicting failures. A person familiar with the art would recognize that other types of tickets may be considered.
[0078] At this point, a time series analysis or study may be undertaken of the data set resulting from the selection of FIG. 11. In one implementation, all event messages for meters which experienced an outage may be listed from oldest to newest in order to facilitate review of when the messages were sent. In one implementation, the time series analysis reveals that most event messages occur in a time period preceding an outage.
[0079] FIG. 12 illustrates an exemplary corresponding histogram or time series plot of AMI event messages. The histogram of FIG. 12 shows a frequency of twelve event or message types occurring between twenty five hours before and five hours after an outage. In this example, the frequency of events increases significantly before an outage. In one implementation, messages occurring within a period ranging from 54 hours before to 6 hours before an outage may be evaluated. The difference between the raw count of events in that 48 hour period may be compared to a corresponding 30 day historical average. In one implementation all AMI events stored in the GreenPlum database and occurring outside of any chosen time frame may be considered as irrelevant for purposes of creating a model for failure prediction.
[0080] In another example, a routine extract may be set up to query the total number of events occurring in the last 24 to 48 hours for comparison against a long term historical average. In one embodiment, the query may be run under the assumption that it will be 50% successful, meaning that 50% of meters which sent an above average (>2x) rate of event messages are likely to experience an outage.
[0081] With the meter events coded, the data may be imported into the data mining engine (for example, IBM's SPSS modeler software). Once inputted into the data mining engine, the data may be transformed such that the event message concentration for each AMI smart meter is calculated. The calculation is shown in equation 1.
Equation 1: Calculating the Concentration of Meter Messages
Concentration of Message A = (# of Message A 54hrs to 6hrs before interruption) / (All
Messages 54hrs to 6 hrs before interruption)
[0082] As an example, suppose meter # 1 outputted 100 total messages between 54 hours before to 6 hours before an interruption occurred. If meter # 1 had 15 'NIC Power Down' event messages in that time period, the concentration of 'NIC Power Down' messages is calculated to be 15%. This calculation may be repeated for all residential and all meter event types. Example 1: Calculating the Concentration of 'Nic Power Down' Messages for Sample
Meter #1
15% = 15 NIC Power Down / 100 Total Messages
[0083] In this example, the message behavior includes a ratio of first messages (15 NIC Power Down messages) over second messages (100 Total Messages), the first messages including a first message type and the second messages including the first message type and at least a second message type, the ratio being 15% which is greater than zero and less than one. Also, it was determined that only 82 events were consistent with electrical messaging, and so the Greenplum AMI event data set was reduced to calculate the message concentrations for these event types.
[0084] A sample of the data prepared for input into the data mining engine is illustrated in FIG. 13. With the raw data input into the data mining engine, an auto-classifier function may be applied to the data, which can be used to determine which models perform the best prediction of a targeted condition. In this example, column A shows the target condition to be an interruption or outage. Column B shows the premises number (in scientific notation) and the remaining columns show twelve of the eighty two event types consistent with electrical messaging and relevant to the outage. Each of the cells associated with the eighty two event types shows a percentage of the concentration of events (shown as a whole number). In one implementation, based on the filtering parameters applied to the data before inputting the data into the data mining engine, prediction models may be selected. In one implementation, when data imbalance exists, either a CHAID (Chi Squared Automatic Interaction Detection) or a C5.1 model may be selected for application to real-time data being received from the meters and predict power failures. In one implementation, the models may be selected based on the analysis of 100 failures in the data set and more than 4.6 million meters without a failure condition. [0085] In one implementation, to ensure modeling is repeatable before selection of the model, 30 or more replicate models may be created and analyzed which may be generated from C5.1, Log Regression, CHAID or other modeling algorithms. In one implementation, the output from the data mining engine may be a decision tree, so, in the example implementation described 30, decision trees from two separate modeling algorithms (C5.1 and CHAID) may be compared for consistency.
[0086] In one implementation, six message concentration types may be considered in order to predict power failures:
Network Interface Card Based Power Fail Detect Disabled History of Direct Current Detected <0.1% concentration
History of Direct Current Detected >0.1% concentration
Network Interface Card Power Down <12.5% concentration Service Error Cleared <1.5% concentration
Zero Voltage Reads
[0087] It should be noted that these six message types are not an exhaustive list. In other examples, other message types and / or other percentage concentrations may be used while remaining within the scope of this description. Also it should be noted that these six message types do not specify any weather information and thus the predictions have the potential advantage of being independent of any weather conditions associated with the power distribution system. An algorithm condition may be set as follows:
Equation 2: Predictive Algorithm Event Message Concentration { NIC Based Power Fail Detect Disabled + (History of Direct Current Detected < 0.1% concentration +) OR (History of Direct Current Detected >0.1% + NIC Power Down < 12.5% concentration)) ( Service Error Cleared < 1.5% concentration) } [0088] A set of SQL queries may be applied so that an automatic data extract of premises meeting the algorithm conditions is executed at 15 minute intervals, in accordance with one embodiment. A sample screenshot of the extraction process is shown in FIG. 14. The results of these queries may be stored in a SQL database where they can be retrieved on demand or according to a script.
[0089] FIG. 15A-FIG. 15C illustrates a decision tree that may be used to predict power failures. In the illustrated implementation, the tree includes tree nodes (as opposed to network nodes) 1501 to 1551. Each node may be associated with a type of event message that is considered to be relevant. The higher the messages appear in the tree structure, the more relevant the message will be in connection with predicting a failure. In the illustrated implementation, the predicted outage or failure is associated with a failure that will result in the generation of a PNC ticket. The following event messages are illustrated: NIC B.P.F., NIC P.D.C.U., Zero V.R., Last Gap, P.F.C., NIC P.D., and NIC F.T.S.
[0090] In the nodes illustrated in FIGs. 15A-C, each node may include a node identification number (for example "Node 1"); a category (for example, "1" means a failure and "0" a no failure); a percentage of confidence level; and the number of premises "n" that had a particular event. Further, the p-value is the measurement of statistical significant level, the chi-square value represents calculated statistics based on chi-square equation, and the "df ' is the degree of freedom - the number of values in the final calculation of a statistic that are free to vary.
[0091] An exemplary "split" is described with respect to nodes 1501 to 1505. The number of premises or locations associated with node 1501 is 10,100 (9999 plus 101). Out of that total number of premises, 47 are identified in node 1503 as having sent a NIC B.P.F. event message (indicated by the > 0.000 notation). Node 1503 further indicates that all 47 premises that sent an NIC B.P.F. message are associated with a "failure" category 1, since the percentage in node 1503 is 100% in connection with the number n=47. What this means is that in relation to the illustrated implementation, the confidence level is 100% - a customer call was generated in connection with all of the 47 premises to report an outage. Therefore, the system of the present disclosure can be used to predict failures before they occur by processing the additional smart meter messages the model of node 1503 to predict a likelihood of an outage at a suspect premise.
[0092] One of the features of the system disclosed herein is that the confidence level threshold can be set to be high, for example, 75% or more, in order to achieve a desired predictability level. For example, in the illustrated implementation, nodes 1503, 1509, 1513, 1523, and 1535 include a confidence level of over 75% in connection with category "1" failure events. These nodes are selected from the multiplicity of end nodes of FIG. 15 (which produce a corresponding multiplicity of prediction models) thus produce selected message behaviors (from a plurality of prediction models 1503, 1509, 1513, 1523, and 1535 selected from the multiplicity of prediction models) for predicting outages (note the message behaviors do not include information specifying any deficient grid element). If the models predict more outages than the repair crew can handle, then the confidence level can be increased and the models re- selected or the results further filtered to match the repair crew capacity. If the models predict fewer outages than the repair crew can handle, then the confidence level can be decreased and the models re- selected or the results less filtered to match the available capacity of the repair crew. For example, the confidence level can be decreased prior to weekend days, seasonal vacation periods or other times when the repair crew has less capacity for repairs so that required repairs may be minimized during those periods when the repair crew has less capacity. [0093] Referring to node 1509, the decision tree shows that out of the total 10100 locations or premises analyzed, the 12 premises that generated both NIC.PD.C.U and Zero V.R. event messages have a probability of 85.714% with respect to reporting a failure. Referring to node 1513, the decision tree shows that out of the total 10100 locations or premises analyzed, the 5 premises that generated both NIC.PD.C.U and Last Gasp event messages have a probability of 100% with respect to reporting a failure. Referring to node 1523, the decision tree shows that out of the total 10100 locations or premises analyzed, the 5 premises that generated both Zero V.R. and P.F.C. event messages have a probability of 100% with respect to reporting a failure. Referring to node 1535, the decision tree shows that out of the total 10100 locations or premises analyzed, the 9 premises that generated NIC.PD event messages at a concentration greater than 0.2% have a probability of 90% with respect to reporting a failure. The decision tree may then be used to identify types of events that may be correlated in order to ascertain whether a single premises will experience a power failure.
[0094] Thus, of the 101 failures of node 1501, 81 failures would have been predicted by selecting the prediction models of nodes 1503 (47 failures), 1509 (14 failures), 1513 (5 failures), 1523 (5 failures), 1535 (10 failures). By using a threshold of a 75% confidence factor for selecting models, a total of 81 failures would have been predicted, 3 of which would have been false detections. More failure could be predicted by including more message behavior models. By repairing the deficient grid elements before a customer generated outage is reported, instead of 101 outage reports generated during the period, only 23 (101-(81-3)) outage reports would have been generated, thereby yielding an over 75% reduction in customer generated outage reports associated with single premises failures and providing a potential advantage of significantly increased customer satisfaction and quality of service. [0095] It should be further noted that end nodes may also be used to validate an absence of an outage. For example, if a message behavior corresponding to node 1539 was determined and yet an outage report was received from associated a suspect premises, then it may be determined that the outage report is erroneous, particularly if the outage report is from a premises generating chronic outage reports. Thus a repair crew would not be dispatched and a "bad roll" would be avoided. In a further example the approach could be applied to a group of premises to determine an absence of outages in a multiplicity of premises within the group of premises.
[0096] Thus, FIG. 8 through FIG. 15 illustrate among other things, a model generator coupled to the message database for generating a multiplicity of models based upon the messages and the outage reports, each model identifying a defective grid element based upon a message behavior of a smart meter associated with the defective grid element, the message behavior including a plurality of messages of a plurality of message types received from the smart meter, and a model selector for selecting a plurality of models from the multiplicity of models based upon a confidence factor.
[0097] The present disclosure has numerous other applications; it may be used to predict failures on other types of devices (single phase laterals or transformers) or in connection with other types of services (neutral legs on a service).
[0098] The foregoing description of possible implementations consistent with the method and system disclosed herein does not represent a comprehensive list of all such implementations or all variations of the implementations described. The description of only some implementation should not be construed as an intent to exclude other implementations. For example, artisans will understand how to implement the system and method disclosed herein in many other ways, using equivalents and alternatives that do not depart from the scope of the system and method disclosed herein. Moreover, unless indicated to the contrary in the preceding description, none of the components described in the implementations are essential to the system and method disclosed herein. It is thus intended that the specification and examples be considered as exemplary only.
[0099] What is claimed is:

Claims

Claims:
1. An electric power distribution system comprising: an electric power distribution grid having a multiplicity of elements capable of becoming deficient;
a multiplicity of smart meters coupling the electric power distribution grid and a corresponding multiplicity of premises, each smart meter measuring electric power consumed at each corresponding premises and generating a multiplicity of message types related thereto, the multiplicity of message types not including information specifying a deficient grid element;
an automated meter infrastructure network coupled to the smart meters for communicating the messages from the smart meters;
a message database coupled to the automated meter infrastructure network for receiving messages from the multiplicity of smart meters;
an outage processor for receiving outage reports through a communication network different from the automated meter infrastructure network, each outage report associated with a failed deficient grid element;
a model generator coupled to the message database for generating a multiplicity of models based upon the messages and the outage reports, each model identifying a defective grid element based upon a message behavior of a smart meter associated with the defective grid element, the message behavior including a plurality of messages of a plurality of message types received from the smart meter;
a model selector for selecting a plurality of models from the multiplicity of models based upon a confidence factor; the message database for receiving additional messages after selection of the plurality of models;
an outage predictor for processing the additional messages with the plurality of models to predict a likelihood of an outage at a suspect premise; and
a ticket generator for generating a ticket enabling a repair crew to repair a deficient grid element associated with the suspect premises.
2. The electric power distribution system according to claim 1 wherein the model generator generates C5.1 and CHAID decision tree models.
3. The electric power distribution system according to claim 1 wherein the model selector determines a confidence level for each of the multiplicity of models and selects the plurality of models based upon the confidence level exceeding a threshold.
4. The electric power distribution system according to claim 3 wherein the confidence level corresponds to 75%.
5. The electric power distribution system according to claim 1 wherein the model generator generates C5.1 and CHAID decision trees having nodes corresponding to models and the model selector determines a confidence level for each of the nodes and selects the plurality of models based upon the confidence level of a corresponding node exceeding a threshold.
6. The electric power distribution system according to claim 1 wherein the model generator scrubs the messages prior to generating the model.
7. The electric power distribution system according to claim 6 wherein the model generator scrubs the messages by at least one of eliminating duplicate messages, selecting messages of a predetermined message type, and selecting messages according to a rate of occurrence of the message.
8. The electric power distribution system of claim 1, wherein the messages include at least one of:
a Network Interface Card Based Power Fail Detect Disabled message;
a History of Direct Current Detected message;
a Network Interface Card Power Down message;
a Service Error Cleared message;
a Zero Voltage Read message; and
a Last Gasp message.
9. A method for maintaining an electric power distribution grid comprising: predicting an outage at a suspect premises based upon a message behavior of a multiplicity of messages generated by a smart meter coupling the electric power distribution grid and the suspect premises; and
generating a repair ticket based upon the predicting.
10. The method according to claim 9 wherein the multiplicity of messages have a multiplicity of message types and the message behavior includes a plurality of the multiplicity of messages having a plurality of the multiplicity of message types.
11. The method according to claim 10 wherein the predicted outage at the suspect premises would result from a deficient element in the electric power distribution grid and the plurality of the multiplicity of messages do not include information specifying the deficient element.
12. The method according to claim 10 wherein the message behavior includes a ratio of first messages over second messages, the first messages including a first message type and the second messages including the first message type and at least a second message type, the ratio being greater than zero and less than one.
13. The method according to claim 10 wherein the message behavior excludes messages of the multiplicity of messages occurring beyond a time period.
14. The method of claim 10 wherein the multiplicity of message types include:
a Network Interface Card Based Power Fail Detect Disabled message;
a History of Direct Current Detected message;
a Network Interface Card Power Down message;
a Service Error Cleared message;
a Zero Voltage Read message; and
a Last Gasp message.
15. The method according to claim 9 wherein the predicting occurs before an occurrence of the outage at the suspect premises.
16. The method according to claim 9 wherein the predicting occurs after an occurrence of the outage at the suspect premises, and the generating the repair ticket occurs before receiving an outage notice associated with the suspect premises, the outage notice being independent of the multiplicity of messages generated by the smart meter.
17. The method according to claim 9 further comprising
determining the message behavior by generating a plurality of prediction models based upon a message database generated by a multiplicity of smart meters associated with a multiplicity of premises and an outage report database of associated with the multiplicity of premises, and
the predicting the outage at the suspect premises analyzes the multiplicity of messages with at least one of the plurality of prediction models to predict a likelihood of the outage at the suspect premises.
18. The method according to claim 17 wherein the generating the plurality of prediction models includes generating a multiplicity of prediction models and selecting the plurality of prediction models from the multiplicity of prediction models based upon a confidence factor.
19. The method according to claim 18 wherein the multiplicity of prediction models are generated based upon at least one of a C5.1 and a CHAID decision tree generation process.
20. The method according to claim 9 wherein the suspect premises is included within a group of premises having a corresponding group of smart meters, the method further comprising predicting an absence of outages at a multiplicity of premises within the group of premises.
21. The method according to claim 20 wherein the multiplicity of premises for which the absence of outages is predicted includes at least ninety nine point nine percent of the group of premises.
22. The method according to claim 21 wherein the group of smart meters generate a big database within a time span, the group of smart meters comprising to at least six hundred thousand smart meters, the group of smart meters generating messages at an average rate of at least two messages per day per smart meter, the time span corresponding to twenty four hours.
23. The method according to claim 9 wherein the predicting is made independent of any weather information specifying any weather condition to which the electric power distribution grid may be exposed.
24. The method according to claim 9 wherein the repair ticket includes information
identifying the suspect premises.
25. The method according to claim 24 wherein the predicted outage at the suspect premises would result from a deficient element in the electric power distribution grid and the repair ticket includes information identifying the deficient element.
26. The method according to claim 9 wherein a non-transitory computer-readable storage media stores instructions for a computer to perform the predicting, and a ticket output device coupled to the computer generates a humanly observable repair ticket indicative of the predicting.
27. A non-transitory computer readable storage medium storing instructions for a computer to perform a method for maintaining an electric power distribution grid, the method comprising:
processing messages generated by a multiplicity of smart meters coupling the electric power distribution grid to a corresponding multiplicity of premises;
processing outage tickets having information regarding outages experienced by at least a portion of the multiplicity of premises, the outage tickets being independent of the messages;
generating a multiplicity of prediction models based upon the processing of the messages and the outage tickets, the prediction models predicting a likelihood of outages based on messages from smart meters; and
selecting a message behavior from the multiplicity of prediction models, the message behavior for predicting an outage at a suspect premises based upon a
multiplicity of messages generated by a smart meter coupling the electric power distribution grid and the suspect premises.
28. The non-transitory computer readable storage medium of claim 27 wherein the predicted outage at the suspect premises would result from a deficient element in the electric power distribution grid and the selected message behavior does not include messages having information specifying the deficient element.
29. The non-transitory computer readable storage medium of claim 27 wherein the
multiplicity of prediction models are generated based upon at least one of a C5.1 and a CHAID decision tree generation process.
30. The non-transitory computer readable storage medium of claim 27 wherein the message behavior includes a plurality messages having a plurality of message types.
31. The non-transitory computer readable storage medium of claim 30 wherein the message behavior includes a ratio of first messages over second messages, the first messages including a first message type and the second messages including the first message type and at least a second message type, the ratio being greater than zero and less than one.
32. The non-transitory computer readable storage medium of claim 30 wherein the message behavior excludes messages of the multiplicity of messages occurring beyond a time period.
33. The non-transitory computer readable storage medium of claim 30 wherein the plurality of message types include at least two of:
a Network Interface Card Based Power Fail Detect Disabled message; a History of Direct Current Detected message;
a Network Interface Card Power Down message;
a Service Error Cleared message;
a Zero Voltage Read message; and
a Last Gasp message. non-transitory computer readable storage medium of claim 30 further comprising: predicting the outage at the suspect premises; and
generating a repair ticket based upon the predicting.
PCT/US2016/027754 2015-06-09 2016-04-15 Outage prevention in an electric power distribution grid using smart meter messaging WO2016200485A1 (en)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US201562173039P 2015-06-09 2015-06-09
US62/173,039 2015-06-09
US15/002,180 2016-01-20
US15/002,180 US20160364648A1 (en) 2015-06-09 2016-01-20 Outage prevention in an electric power distribution grid using smart meter messaging

Publications (1)

Publication Number Publication Date
WO2016200485A1 true WO2016200485A1 (en) 2016-12-15

Family

ID=57504584

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2016/027754 WO2016200485A1 (en) 2015-06-09 2016-04-15 Outage prevention in an electric power distribution grid using smart meter messaging

Country Status (2)

Country Link
US (2) US20160364648A1 (en)
WO (1) WO2016200485A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110110924A (en) * 2019-05-05 2019-08-09 国网经济技术研究院有限公司 A kind of power distribution network scale of investment optimization method, device and calculate equipment

Families Citing this family (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11816465B2 (en) 2013-03-15 2023-11-14 Ei Electronics Llc Devices, systems and methods for tracking and upgrading firmware in intelligent electronic devices
US11734396B2 (en) 2014-06-17 2023-08-22 El Electronics Llc Security through layers in an intelligent electronic device
US10958435B2 (en) 2015-12-21 2021-03-23 Electro Industries/ Gauge Tech Providing security in an intelligent electronic device
KR101789923B1 (en) * 2016-06-02 2017-10-25 두산중공업 주식회사 Wind farm supervision monitoring method
US10191506B2 (en) * 2016-09-29 2019-01-29 Enel X North America, Inc. Demand response dispatch prediction system including automated validation, estimation, and editing rules configuration engine
US10185728B2 (en) * 2016-12-19 2019-01-22 Capital One Services, Llc Systems and methods for providing data quality management
US11107045B2 (en) 2017-03-20 2021-08-31 Florida Power & Light Company Smart meter based service ticket closing control
CN107292414B (en) * 2017-05-02 2021-02-09 国网浙江省电力有限公司 Resident user distributed internet optimal selection method based on peak-valley electric quantity prediction
US11686594B2 (en) 2018-02-17 2023-06-27 Ei Electronics Llc Devices, systems and methods for a cloud-based meter management system
US11754997B2 (en) 2018-02-17 2023-09-12 Ei Electronics Llc Devices, systems and methods for predicting future consumption values of load(s) in power distribution systems
US11734704B2 (en) * 2018-02-17 2023-08-22 Ei Electronics Llc Devices, systems and methods for the collection of meter data in a common, globally accessible, group of servers, to provide simpler configuration, collection, viewing, and analysis of the meter data
US11539609B2 (en) * 2018-09-11 2022-12-27 Trilliant Networks, Inc. Method and apparatus for reporting power down events in a network node without a backup energy storage device
CN110083644B (en) * 2019-03-04 2023-04-28 广东电网有限责任公司 Method and system for early warning of major outage event of distribution network line
CN110188122B (en) * 2019-05-15 2021-12-10 电子科技大学 Incidence relation analysis method for different line loss behaviors
US11863589B2 (en) 2019-06-07 2024-01-02 Ei Electronics Llc Enterprise security in meters
CN110795837B (en) * 2019-10-21 2023-03-03 国网安徽省电力有限公司安庆供电公司 Automatic distribution network ticket application one-time mode repeating operation method based on topological linkage
US11327938B2 (en) * 2019-11-22 2022-05-10 Sap Se Method to improve prediction accuracy of business data with enhanced data filtering and data reduction mechanism
CN111382501B (en) * 2020-02-21 2023-06-02 华北电力大学 Configuration method and device of intelligent ammeter and computing equipment
US10904778B1 (en) 2020-03-04 2021-01-26 Cisco Technology, Inc. Detection and deduction of urgent messages in low power and lossy networks
CN111564903B (en) * 2020-05-27 2022-04-29 中国电力科学研究院有限公司 Intelligent distribution transformer terminal, intelligent electric meter access method and distribution Internet of things

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050001009A1 (en) * 2001-08-11 2005-01-06 Klaus Raether Friction stir tool for friction welding
US20090001012A1 (en) * 2005-11-03 2009-01-01 Bryan Kepner Substrates Comprising Anti-Microbial Compositions and Methods of Making and Using the Same
US7966150B2 (en) * 2005-11-17 2011-06-21 Florida Power & Light Company Data analysis applications
US20120200423A1 (en) * 2011-02-08 2012-08-09 Avista Corporation Outage Prediction With Next Generation Smart Grid
US20130066570A1 (en) * 2011-09-13 2013-03-14 General Electric Comapny Selection of bellwether smart grid meters
US20140114829A1 (en) * 2012-10-24 2014-04-24 Joseph W. Forbes, Jr. System, method, and apparatus for settlement for participation in an electric power grid
US8779927B2 (en) * 2010-09-07 2014-07-15 Grid Net, Inc. Power outage notification
US20150051749A1 (en) * 2011-12-16 2015-02-19 Schneider Electric Usa Inc. Co-location electrical architecture

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US7089125B2 (en) 2003-10-27 2006-08-08 Itron, Inc. Distributed asset optimization (DAO) system and method
US20060117059A1 (en) * 2004-11-30 2006-06-01 Tidal Software, Inc. System and method for monitoring and managing performance and availability data from multiple data providers using a plurality of executable decision trees to consolidate, correlate, and diagnose said data
US20090125155A1 (en) 2007-11-08 2009-05-14 Thomas Hill Method and System for Optimizing Industrial Furnaces (Boilers) through the Application of Recursive Partitioning (Decision Tree) and Similar Algorithms Applied to Historical Operational and Performance Data
WO2009067250A1 (en) 2007-11-25 2009-05-28 Trilliant Networks, Inc. System and method for false alert filtering of event messages within a network
US8892375B2 (en) * 2008-05-09 2014-11-18 Accenture Global Services Limited Power grid outage and fault condition management
US8924033B2 (en) * 2010-05-12 2014-12-30 Alstom Grid Inc. Generalized grid security framework
US20130338945A1 (en) 2012-06-13 2013-12-19 Abb Research Ltd. Outage Scope Analysis for Electrical Distribution Systems
US9667521B2 (en) * 2014-01-27 2017-05-30 Vencore Labs, Inc. System and method for network traffic profiling and visualization

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050001009A1 (en) * 2001-08-11 2005-01-06 Klaus Raether Friction stir tool for friction welding
US20090001012A1 (en) * 2005-11-03 2009-01-01 Bryan Kepner Substrates Comprising Anti-Microbial Compositions and Methods of Making and Using the Same
US7966150B2 (en) * 2005-11-17 2011-06-21 Florida Power & Light Company Data analysis applications
US8779927B2 (en) * 2010-09-07 2014-07-15 Grid Net, Inc. Power outage notification
US20120200423A1 (en) * 2011-02-08 2012-08-09 Avista Corporation Outage Prediction With Next Generation Smart Grid
US20130066570A1 (en) * 2011-09-13 2013-03-14 General Electric Comapny Selection of bellwether smart grid meters
US20150051749A1 (en) * 2011-12-16 2015-02-19 Schneider Electric Usa Inc. Co-location electrical architecture
US20140114829A1 (en) * 2012-10-24 2014-04-24 Joseph W. Forbes, Jr. System, method, and apparatus for settlement for participation in an electric power grid

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110110924A (en) * 2019-05-05 2019-08-09 国网经济技术研究院有限公司 A kind of power distribution network scale of investment optimization method, device and calculate equipment

Also Published As

Publication number Publication date
US20160364648A1 (en) 2016-12-15
US11062212B2 (en) 2021-07-13
US20200151581A1 (en) 2020-05-14

Similar Documents

Publication Publication Date Title
US11062212B2 (en) Outage prevention in an electric power distribution grid using smart meter messaging
Bhusal et al. Power system resilience: Current practices, challenges, and future directions
Kandaperumal et al. Resilience of the electric distribution systems: concepts, classification, assessment, challenges, and research needs
CN104268678B (en) A kind of petrochemical equipment preventative maintenance method based on dynamic reliability
US9652723B2 (en) Electrical transformer failure prediction
AU2013257529B2 (en) Monitoring system
JP5452613B2 (en) Power grid supply interruption and failure status management
Yen et al. Effect of smart meter data collection frequency in an early detection of shorter-duration voltage anomalies in smart grids
US20090187344A1 (en) System, Method, and Computer Program Product for Analyzing Power Grid Data
Baldick et al. Vulnerability assessment for cascading failures in electric power systems
JP2018061425A (en) Enhanced disturbance management of power grid system
Di Giorgio et al. A Bayesian network-based approach to the critical infrastructure interdependencies analysis
CN104487991A (en) Energy management with correspondence based data auditing signoff
EP3615892B1 (en) Method and system for detecting anomalies in energy consumption
CN105606920A (en) System and method for determining the current and future state of health of a power transformer
Goyal et al. Asset health management using predictive and prescriptive analytics for the electric power grid
Perdue et al. Energy yields of small grid connected photovoltaic system: effects of component reliability and maintenance
JP2013538543A (en) Machine learning for power grids
Stringer et al. Observed behavior of distributed photovoltaic systems during major voltage disturbances and implications for power system security
Nan et al. Multilayer hybrid modeling framework for the performance assessment of interdependent critical infrastructures
Dehghani et al. An overview of the assessment metrics of the concept of resilience in electrical grids
Amini et al. Electrical energy systems resilience: A comprehensive review on definitions, challenges, enhancements and future proceedings
Sun et al. A framework for dynamic prediction of reliability weaknesses in power transmission systems based on imbalanced data
CN105137776A (en) Metering automation terminal control assistant system
Guo Data analytics and application developments based on synchrophasor measurements

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16807958

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16807958

Country of ref document: EP

Kind code of ref document: A1